Read more

Powerful A.I. Is Coming. We’re Not Ready. (NYT)

Getting divorced? Artificial intelligence deepfakes could cost you in court (Fox News)

How the UK tech secretary uses ChatGPT for policy advice (newscientist.com)

*Paige works for GitHub, a Microsoft Company

Dashcam’s Binary: Safety or Privacy?

AI-powered dashcams are revolutionizing the trucking industry by enhancing road safety and fleet management. These devices use advanced computer vision to detect unsafe driving behaviors, such as rolling stops, lane changes, driver drowsiness, and seatbelt violations. Companies like FusionSite Services have reported significant safety improvements, including an 89% reduction in accidents. However, the widespread use of AI dashcams raises privacy concerns for truck drivers, who often view their vehicles as personal spaces, and for other motorists recorded without consent.

👉 Why it matters: This use of facial recognition technology is a double-edged sword. It could be used to identify trafficking victims or identify the license plates of cars that are on the run or party to an Amber Alert. However, legal challenges, such as lawsuits over biometric data collection, highlight the need for compliance with privacy regulations. To balance safety with privacy, companies are implementing features like Driver Privacy Mode and data security measures. Open communication and clear policies can help drivers adapt to these changes. The trucking industry must navigate this evolving landscape carefully, ensuring that safety benefits do not come at the cost of individual privacy rights.

Responsible AI rating: 🟡

Android Could Capitalize on Apple’s Big AI Mishaps

Apple’s AI ambitions have hit a rough patch, as internal turmoil over Siri's shortcomings and delayed Apple Intelligence features cast a shadow over its future. Apple executives have privately admitted to an "ugly and embarrassing" rollout, with AI-powered features working inconsistently and missing key release deadlines. Siri, already seen as lagging behind AI competitors, now risks falling even further behind as Apple scrambles to fix fundamental flaws.

Meanwhile, Google is pushing forward with new AI-powered features in Android 16, including AI-generated notification summaries for messaging apps. Unlike Apple's failed attempt at news summaries—which delivered misinformation—Google's more cautious approach may help Android gain an edge in AI-driven user experiences. While Apple’s AI delays extend into 2025, Android is poised to deliver practical, AI-powered enhancements that actually work.

With Apple struggling to integrate AI seamlessly into its ecosystem, Android’s steady progress could shift the balance, proving that AI innovation is only valuable if it delivers real-world benefits without undermining trust. As Apple stumbles, Android has a rare opportunity to redefine itself as the leader in mobile AI.

Responsible AI rating: 🟢

👉 Why it matters: This matters because it shows the competition in the AI space is alive and well, which could be good news for responsible AI. Companies are now breaking out of the “we have to chase other companies to catch up” mindset and are focused on charting their own paths. Apple has pushed their updates in favor of creating something truly useful rather than launching vaporware. Android is going to try to do what Apple did with news alerts - but better. Competition promotes choice.

China’s Manus Agent: What it is and why it matters

Chinese startup Butterfly Effect has unveiled Manus, an AI agent that operates autonomously without requiring step-by-step human instructions. Unlike traditional chatbots like ChatGPT, Manus can break down and complete complex tasks on its own using multiple AI models. Early users have tested its capabilities by creating video games, designing websites, and analyzing financial data. However, the AI faces challenges, including crashes and occasional misunderstanding of tasks. Its development hints at the potential future of artificial general intelligence (AGI).
Responsible AI rating: 🟡

👉 Why it matters: While Manus shows promise in advancing AI autonomy, it also raises ethical concerns about control and oversight. The major challenge with this technology is that it’s another product that’s not broadly available, that’s boasting to be the future of artificial general intelligence (AGI). AGI, of course, is controversial, because everyone is essentially making up their own definitions. More and more AI companies will boast about achieving AGI, but they’ll do that without product proof and with the goal of making money. It’s important that we all stay sharp and critically evaluate these new AI uses.

Learn More

Keeping Watch: US State AI Laws

Illinois and California are taking steps to regulate AI, particularly in healthcare and high-impact decision-making. In Illinois, lawmakers are targeting AI’s role in mental health services and health insurance decisions. House Bill 1806 would prohibit AI from assisting in therapy sessions, citing ethical concerns and the inability of AI to recognize crises. House Bill 35, meanwhile, seeks to prevent health insurers from solely relying on AI to deny or reduce coverage, requiring human oversight. The legislation follows reports of AI-driven claim denials and lawsuits against major insurers for using automated systems to reject care.

In California, lawmakers are advancing a broader set of AI regulations, with Assembly Bill 1018 aiming to prevent automated discrimination in employment, education, healthcare, housing, and finance. The bill mandates AI performance evaluations and transparency when AI makes consequential decisions. Additionally, California is revisiting previous AI regulatory efforts, including protections for AI whistleblowers and oversight of AI’s impact on government services.

👉 Why it matters: The push for state-level AI regulation comes as federal oversight remains uncertain, especially after the Trump administration rescinded Biden-era AI protections. With major tech companies lobbying against strict AI rules, states like Illinois and California are positioning themselves as leaders in AI accountability, seeking to balance innovation with consumer protection. One challenge is that AI companies are trying to get relief from these laws in hopes that Federal laws will be more tailored to what they want, and will allow them to innovate in the way that they see fit.

China Mandates Labels for AI-Generated Content

China will require AI-generated content to be clearly labeled under new regulations effective September 1, aiming to curb misinformation and enhance transparency. Issued by the Cyberspace Administration of China (CAC) and other government agencies, the rules mandate both visible labels and embedded digital watermarks for AI-generated text, images, audio, and video. Online platforms must verify AI content before publication and retain records for six months. Removing or altering labels is strictly prohibited. The move aligns with Beijing’s broader AI oversight push, mirroring global efforts like the EU’s AI Act. However, challenges remain in enforcing real-time AI applications and watermark reliability.

👉 Why it matters: AI is influencing the market across the world. Deepfakes, in particular, are causing market havoc because people are drumming up false catastrophes, fooling people, and it’s causing people to react by selling shares. With this in mind, China is mandating labels for AI generated content in order to reduce the market churn and increase transparency.

The AI Action Plan Submissions are In

​In February 2025, the White House Office of Science and Technology Policy (OSTP) issued a Request for Information (RFI) seeking public input to develop an Artificial Intelligence (AI) Action Plan aimed at sustaining and enhancing America's global AI leadership. The RFI invited feedback from various stakeholders, including academia, industry groups, private sector organizations, and government entities, with a submission deadline of March 15, 2025.​

Several prominent organizations have submitted their recommendations:

• Google: Emphasized investing in AI infrastructure, modernizing government AI adoption, and promoting pro-innovation approaches internationally.

• OpenAI: Proposed strategies to ensure freedom to innovate, export democratic AI, promote the freedom to learn, and capitalize on infrastructure opportunities to drive growth.

• Anthropic: Focused on preparing for the emergence of powerful AI systems by advocating for responsible scaling policies and robust testing and evaluation frameworks. ​

• R Street Institute: Highlighted the importance of a pro-innovation policy framework to maintain U.S. competitiveness and address potential risks associated with AI deployment.

• Premier Inc.: Discussed leveraging AI to reduce healthcare costs, improve quality and access, and alleviate workforce shortages, while emphasizing data standards and responsible data use. ​

With the public comment period concluded, the OSTP will review the submissions to inform the development of the AI Action Plan. This plan is expected to outline priority policy actions to bolster America's AI capabilities, promote innovation, and address potential challenges associated with AI technologies. A finalized version of the plan is anticipated to be presented to the President by mid-October 2025. ​

👉 Why it matters: These recommendations will shape the future of AI policy in the US. What remains to be seen is whether the OSTP will keep an open mind to the wide variety of perspectives and recommendations shared, or if they’re simply on the lookout for novel ideas that already align with their pre-determined AI policy agenda.

Spotlight on Research

• Balancing Practical Uses and Ethical Concerns:The Role of Large Language Models in Scientific Research
  The rapid adoption of artificial intelligence (AI) in scientific research is accelerating progress but also challenging core scientific norms such as accountability, transparency, and replicability. Large language models (LLMs) like ChatGPT are revolutionizing scientific communication and problem-solving, but they introduce complexities regarding authorship and the integrity of scientific work. LLMs have the potential to transform various research practices, including literature surveys, meta-analyses, and data management tasks like entity resolution and query synthesis. Despite their advantages, LLMs present challenges such as content verification, transparency, and accurate attribution. This study explores the appropriate use of LLMs for NASA's Science Mission Directorate (SMD), considering whether to develop a custom bespoke model or fine-tune an existing open-source model. This article reviews the outcomes and lessons learned from this effort, providing insights for other research groups navigating similar decisions.

• The Unified Control Framework: Establishing a Common Foundation for Enterprise AI Governance, Risk Management and Regulatory Compliance

  The rapid adoption of AI systems presents enterprises with a dual challenge: accelerating innovation while ensuring responsible governance. Current AI governance approaches suffer from fragmentation, with risk management frameworks that focus on isolated domains, regulations that vary across jurisdictions despite conceptual alignment, and high-level standards lacking concrete implementation guidance. This fragmentation increases governance costs and creates a false dichotomy between innovation and responsibility. We propose the Unified Control Framework (UCF): a comprehensive governance approach that integrates risk management and regulatory compliance through a unified set of controls. The UCF consists of three key components: (1) a comprehensive risk taxonomy synthesizing organizational and societal risks, (2) structured policy requirements derived from regulations, and (3) a parsimonious set of 42 controls that simultaneously address multiple risk scenarios and compliance requirements. We validate the UCF by mapping it to the Colorado AI Act, demonstrating how our approach enables efficient, adaptable governance that scales across regulations while providing concrete implementation guidance. The UCF reduces duplication of effort, ensures comprehensive coverage, and provides a foundation for automation, enabling organizations to achieve responsible AI governance without sacrificing innovation speed.

  ---

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai.

Connect with me on LinkedIn, or subscribe to the Just AI with Paige YouTube channel to see more videos on responsible AI.

Looking for a speaker at your next AI event? Email hello@justaimedia.ai if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles. I always read the summary to check that it’s factual and accurate.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

If you prefer not to read the newsletter, you can listen instead!👂

Spotify

Apple

Who bought this smoked salmon? How ‘AI agents’ will change the internet (and shopping lists)(The Guardian)

What one Finnish church learned from creating a service almost entirely with AI (AP News)

Meet the 21-year-old helping coders use AI to cheat in Google and other tech job interviews (CNBC)

*Paige Lord is the founder & CEO of Just AI. She also works for GitHub, a Microsoft Company

Nuclear: Achieving net-zero emissions as AI energy demand rises

Microsoft is investing $1.6 billion to restart the Three Mile Island nuclear plant, leveraging nuclear power to meet AI data centers' growing energy needs. Big Tech companies like Google and Amazon are also turning to nuclear energy to achieve net-zero emissions, with Google investing in small modular reactors (SMRs). Experts warn that SMRs are still unproven, and tech companies may struggle to scale them. Despite uncertainties, industry leaders believe nuclear could be key to meeting clean energy goals amid rising AI electricity demands. (Read the article.)

👉 Why it matters: As energy demand for AI products increase, companies are scrambling to balance their commitments to sustainability and their commitment to shareholders. This has led to the rapid exploration of energy options that will allow companies to build AI, power it, and meet demand for the products, but also help companies uphold their voluntary - and crucial - commitments to the future of the planet. While nuclear power is reliable and carbon-free, challenges include high costs, long build times, and nuclear waste. This begs the question: is there any good, clean, non-detrimental option for powering AI?

Responsible AI rating: 🟡

Apple’s speech-to-text is insulting - literally

A Scottish woman, Louise Littlejohn, was shocked when Apple’s AI-powered voice-to-text transcription mistakenly inserted an inappropriate reference to sex and an insult in a voicemail from a Land Rover dealership. The message was a routine business call inviting her to an event, but the AI system seemingly misheard words, likely due to background noise and the scripted nature of the call. Experts believe the Scottish accent may have contributed, but poor audio quality was a bigger factor. Apple declined to comment, but similar AI transcription issues have been reported before, highlighting concerns about speech-to-text accuracy and safeguards in public AI systems. (Read the article.)

👉 Why it matters: As AI continues to scale and more use cases are identified, there is an expectation that there will be some errors. However, speech-to-text technology has been around for some time. This, coupled with Apple’s other AI-related challenges, like the spreading of misinformation in the form of AI hallucinations on iOS devices, is a cause for concern and further illustrates Apple’s race to release AI may have left them on the back foot. However, Apple has a good track record of pulling their AI off the market when there are serious errors, and working to release safer, less problematic technology.

Responsible AI rating: 🟡

Learn More

Apple pushes AI-enhanced Siri back to 2026

Apple has delayed its AI-enhanced Siri features until 2026, pushing back capabilities that would allow Siri to interact with other apps and use personal context for tasks like filling out forms. Originally expected this spring, the delay underscores Apple's struggles to keep pace with AI rivals like OpenAI, Amazon, and Google. Apple has already integrated some AI-driven Siri improvements, including a more conversational tone and ChatGPT integration. However, Apple Intelligence has faced challenges, including inaccuracies in AI-generated news summaries. Developers are preparing Siri-compatible features, but they won’t be functional until Apple releases a beta. Apple is expected to announce updates at WWDC in June. (Read the article.)

👉 Why it matters: Apple’s already seeing challenges with their Apple Intelligence platform and features after a hasty push to get their AI products in market. This setback is another sign that Apple isn’t in a place to compete well with other AI companies, and it begs the question: do they have a vision? Ultimately, it’s better to launch a higher quality, safer AI product on a longer timeline than a poor quality, unsafe product just to say you’re in the market.

Responsible AI rating: 🟢

The US State Department is using AI to target foreign student visa holders participating in protests, chills free speech

The State Department's new "Catch and Revoke" initiative uses AI technology to scan tens of thousands of foreign student visa holders’ social media accounts for signs of support for Hamas or other U.S.-designated terror groups. The AI will also flag participation in anti-war or pro-Palestinian demonstrations, checking news reports and lawsuits citing antisemitic incidents. Officials say the program is part of a "whole of government" effort involving the Justice and Homeland Security Departments.

Critics, including civil rights groups, argue the program raises serious First Amendment concerns and could encourage self-censorship on college campuses. The Foundation for Individual Rights and Expression warned that AI is not equipped to judge protected speech, while others likened the initiative to past surveillance programs targeting pro-Palestinian activists.

The move aligns with President Trump’s executive orders, which aim to deport non-citizen students who participate in pro-Palestinian protests. Trump recently pulled $400 million in federal funding from Columbia University over alleged failures to protect Jewish students. Critics fear AI’s role in policing speech could lead to broad overreach, affecting free expression on U.S. campuses. (Read the article.)

👉 Why it matters: Freedom of speech and freedom to assemble are two rights afforded by the First Amendment for citizens and residents of the United States.

“Congress shall make no law respecting an establishment of religion, or prohibiting the free exercise thereof; or abridging the freedom of speech, or of the press; or the right of the people peaceably to assemble, and to petition the Government for a redress of grievances.”

This use of AI by the United States government catapults us into dangerous territory as a country. First, because it has a chilling effect on free speech and the right to peaceably assemble. Second, because this use case is being justified by a false narrative. The State Department has indicated that they believe that if someone stands with the Palestinian people, they also stand with terrorist groups like Hamas. This change of narrative, which advances the President’s agenda of seizing Gaza and turning it into a resort, is meant to create a false link between the two. Third, the use of AI to scan social media to identify student protesters as a way to “catch” the students and “revoke” their student visas is dangerous. Facial recognition technology and other vision AI used to examine these social media accounts and flag “offenders” is a road to harm. The Justice Department is not being transparent about the type of AI being use, safety built in, or how they have humans incorporated into the process to ensure that people are not being falsely identified.

China welcomes AI expert home after two decades abroad

Chinese AI expert Tingwen Huang has returned to China after over two decades abroad, joining Shenzhen University of Advanced Technology as a chair professor. Previously, he spent 20+ years at Texas A&M University in Qatar, where his research focused on intelligent control, optimization, and complex systems dynamics. Huang is a highly cited researcher, with nearly 700 published papers and over 44,000 citations on Google Scholar. His work, supported by $7 million in funding from the Qatar National Research Fund, applies mathematical principles to AI, neural networks, and autonomous systems. (Read the article.)

👉 Why it matters: Huang’s return signifies China's efforts to attract top AI talent as it strengthens its position in AI research and development. This is the latest in a series of signals that indicate that China is continuing to ramp up their AI efforts. Last week, Chinese authorities issued a travel warning to AI entrepreneurs and researchers advising them to limit their travel to the US out of fear that there could be potential intelligence leaks and strategic vulnerabilities. China also quietly removed their first minister of industry and technology - Jin Zhuanglong - and replaced him with another leader. Finally, China has been expanding their investment in “The Digital Silk Road” - implementing digital infrastructure in Africa. They’ve signed over $700 billion in engineering deals in Africa in the last decade, and are now expanding AI infrastructure as well.

Anthropic shares policy recommendations for the US AI Action Plan

Anthropic has submitted key recommendations to the Office of Science and Technology Policy (OSTP) for the U.S. AI Action Plan, urging decisive action to maintain America’s AI leadership. The company anticipates powerful AI systems emerging by 2026-2027, with human-like intellectual capabilities and autonomous control over digital and physical tools.

Anthropic’s six key policy proposals include:

• National security testing of AI models.

• Stronger export controls on semiconductor technology.

• Enhanced AI lab security with classified intelligence coordination.

• Expanding energy infrastructure by 50 gigawatts by 2027.

• Accelerating AI adoption across government agencies.

• Preparing for economic shifts caused by AI advances.

These recommendations aim to balance innovation with security risks, ensuring AI benefits all Americans. (Read the article.)

👉 Why it matters: As the US continues to work out its stance on AI policy and is collecting public comments, we’re seeing more companies throw their perspectives in the ring. Anthropic’s suggestions is the latest in a series of large AI company recommendations on AI policy. In contrast to what companies like Microsoft are saying, Anthropic is asking for stronger export controls on semiconductor technology, which shows a stark difference in perspectives. As perspectives are taken into account, it will be interesting to know which company’s suggestions have sticking power with the Trump administration.

Spotlight on Research

• Generative AI, Democracy and Human Rights

  Disinformation is not new, but given how disinformation campaigns are constructed, there is almost no stage that will not be rendered more effective by the use of generative artificial intelligence (AI). Given the unsatisfactory nature of current tools to address this budding reality, disinformation, especially during elections, is set to get much, much worse.

  As these campaigns become more sophisticated and manipulative, the foreseeable consequence will be a further erosion of trust in institutions and a heightened disintegration of civic integrity, which in turn will jeopardize a host of human rights, including electoral rights and the right to freedom of thought.

  In this policy brief, David Evan Harris and Aaron Shull argue that policy makers must hold AI companies liable for the harms caused or facilitated by their products that could have been reasonably foreseen, act quickly to ban using AI to impersonate real persons or organizations, and require the use of watermarking or other provenance tools to allow people to distinguish between AI-generated and authentic content.

  ---

Watch: Frontier Insights with host Yogesh Chavda

Watch on LinkedIn!

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai.

Connect with me on LinkedIn, or subscribe to the Just AI with Paige YouTube channel to see more videos on responsible AI.

Looking for a speaker at your next AI event? Email hello@justaimedia.ai if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles. I always read the summary to check that it’s factual and accurate.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

Check out our Just AI Weekly Update podcast! We cover the top AI, responsible AI, and AI policy news stories of the week - and why it matters - in 25 minutes or less.

Listen on Spotify

Listen on Apple

Listen on YouTube

Anthropic debuts Claude 3.7 Sonnet with deep reasoning (Axios)

DeepSeek rushes to launch new AI model as China goes all in (Reuters)

Microsoft Dropped som AI Data Center Leases (Bloomberg)

*Paige works for GitHub, a Microsoft Company

Microsoft Debuts the Majorana 1 Chip, Potentially Revolutionizing Quantum Computing

Microsoft has unveiled Majorana 1, a groundbreaking quantum chip based on a Topological Core architecture. Utilizing a novel material called a topoconductor, it enables more stable and scalable qubits, a major step toward quantum computers with a million qubits. This breakthrough promises to solve industrial-scale problems, such as environmental cleanup and material science innovations, within years rather than decades. The chip’s digital control simplifies quantum operations, improving reliability and efficiency. Microsoft’s unique approach, validated by peer-reviewed research in Nature, led to DARPA’s endorsement for utility-scale quantum computing. By leveraging Majorana particles—previously theoretical—Microsoft has achieved a key milestone in fault-tolerant quantum computing. The company aims to integrate these advancements into Azure Quantum, providing scalable solutions for commercial and scientific applications. This progress could revolutionize AI, chemistry, and engineering, making quantum computing an essential tool for future problem-solving.

For a deeper dive on this topic, I recommend checking out Meg McNulty’s Cipher Talk deep dive article - Microsoft’s Quantum Leap.

👉 Why it matters: Risks related to this breakthrough include the typical understanding that any technology can be a tool or a weapon. So it is with AI, so it is with quantum computing. As the technology scales and becomes less expensive in the future, we could see a reality where bad actors use a quantum/AI combination to achieve catastrophic ends. One hope is that the AI era gives enough runway to land on realistic and meaningful safety guidelines that can be applied to quantum innovations. Furthermore, as with AI, impact on the planet should be considered, especially given that qubits must be kept at absolute zero (-460 degrees Fahrenheit), and the energy required to achieve that is tremendous.

Responsible AI rating: 🟡

xAI Grok 3 vs. OpenAI o3-Mini-High AI Benchmark Controversy

A dispute over AI benchmarking erupted after an OpenAI employee accused xAI of misleading claims about its Grok 3 model’s performance on AIME 2025, a math benchmark. xAI’s chart showed Grok 3 outperforming OpenAI’s o3-mini-high but omitted the “cons@64” metric, which boosts scores by allowing multiple attempts. When using this metric, OpenAI’s models performed better. xAI co-founder Igor Babuschkin defended the company, noting OpenAI has used similar tactics. Experts argue AI benchmarks often lack transparency, especially regarding computational costs. This debate highlights the ongoing challenges in fairly assessing AI model capabilities and communicating their real-world effectiveness.

👉 Why it matters: AI benchmark results can be misleading if not critically examined. While they offer insight into model performance, they are often arbitrary and don’t fully capture a model’s real-world capabilities or limitations. As AI researcher Nathan Lambert points out, the most important but often overlooked factor is the computational and financial cost required to achieve top scores. Without transparency on these aspects, comparisons between models can be skewed. It’s crucial to fact-check claims, understand the context behind the numbers, and recognize that benchmarks alone don’t determine an AI model’s true effectiveness.

Responsible AI rating: 🚩

Learn More

Microsoft Employees Protest Microsoft’s Cloud Contracts with the Israeli Military

Five Microsoft employees were ejected from a company town hall after protesting Microsoft’s AI and cloud contracts with the Israeli military. Their protest followed an AP investigation revealing that Microsoft's AI models were used in Israeli military targeting during conflicts in Gaza and Lebanon. The employees wore shirts spelling “Does Our Code Kill Kids, Satya?” before being removed. Microsoft has faced internal dissent on this issue for months, with some employees questioning whether these contracts violate the company’s human rights commitments. The company has not disclosed whether the protesters will face disciplinary action, but it previously fired two workers for organizing a vigil for Palestinian refugees.

👉 Why it matters: This incident highlights growing tensions within major tech firms over ethical AI use and military contracts. As AI becomes more integrated into warfare, companies like Microsoft face scrutiny over their role in global conflicts. Employee activism on ethical concerns is increasing, raising questions about corporate responsibility, transparency, and alignment with stated human rights principles. The controversy also underscores the broader debate on AI’s ethical deployment, particularly in life-or-death military applications.

Responsible AI rating: 🟡

Open AI Uncovers Evidence of Chinese Surveillance Tool

OpenAI uncovered evidence of a Chinese AI-powered surveillance tool designed to track anti-Chinese posts on Western social media. Researchers identified the tool, dubbed Peer Review, after someone used OpenAI’s technology to debug its code. OpenAI also found another Chinese campaign, Sponsored Discontent, which generated English-language posts criticizing Chinese dissidents and translated articles into Spanish to spread anti-U.S. narratives. Additionally, a Cambodia-based operation used OpenAI’s models to aid online scams. These findings highlight growing concerns about AI’s role in surveillance, disinformation, and cybercrime, while also demonstrating how AI can be leveraged to detect and counter such threats.

👉 Why it matters: This case highlights the importance of being mindful when using AI. OpenAI’s ability to track how its technology is used underscores that AI companies are monitoring activity—sometimes for good, but not always. Businesses should be especially cautious, as sharing sensitive information with AI tools could lead to unintended exposure.

Additionally, this incident shows AI’s potential for positive impact. OpenAI’s detection of misuse provided valuable insights into China’s disinformation tactics and surveillance efforts.

Finally, it’s a reminder to educate loved ones about online scams. Techniques like pig butchering have led to devastating financial losses—awareness is key to prevention.

Political Consultant who Commissioned Biden DeepFake Calls Goes to Trial

Democratic consultant Steve Kramer is on trial for orchestrating AI-generated robocalls mimicking President Biden ahead of the 2024 New Hampshire primary. Kramer faces 26 criminal charges and a $6 million FCC fine for attempting to mislead voters. The calls, reaching up to 25,000 people, falsely urged Democrats to skip the primary. Kramer claims it was a stunt to highlight AI regulation needs. A magician was paid $150 to create the deepfake. Lingo Telecom, which transmitted the calls, faces a $2 million fine. New Hampshire officials launched an investigation, underscoring concerns over AI’s role in election misinformation.

👉 Why it matters: This case highlights the dangerous potential of AI in spreading election misinformation. While Kramer claims his actions were meant to expose the need for AI regulation, the impact remains deeply harmful to democracy. AI-generated deepfakes can erode trust in elections, manipulate voters, and create confusion, setting a troubling precedent. As AI technology advances, stronger safeguards are needed to prevent similar incidents from undermining democratic processes and public confidence in fair elections. Intentions aside, the misuse of AI in this way demonstrates the urgent need for ethical and legal protections.

EU Withdraws AI Liability Proposal

The EU has withdrawn the AI Liability Directive, which aimed to protect victims of AI-related harm, from its 2025 work program due to stalled negotiations. However, the European Parliament’s Internal Market and Consumer Protection Committee is pushing to continue work on it, while the Legal Affairs Committee has yet to decide. Critics, including MEP Axel Voss, blame industry lobbying for the reversal, arguing that Big Tech fears liability rules. Meanwhile, in the U.S., the White House is seeking public input on its AI Action Plan, which will shape policies on AI governance, security, innovation, and regulation, with comments open until March 15, 2025.

👉 Why it matters: If the AI Liability Directive is indefinitely shelved, individuals harmed by AI systems will have fewer legal protections and limited recourse for proving harm. Without liability rules, the EU would lack the necessary transparency requirements for AI companies, making it harder for victims to seek justice.

The decision may also be influenced by discussions at the Paris AI Action Summit, where the EU acknowledged that regulatory complexity and red tape were deterring businesses. While streamlining AI governance is important, removing liability protections risks prioritizing corporate interests over consumer rights and accountability.

Spotlight on Research

• "Why do we do this?": Moral Stress and the Affective Experience of Ethics in Practice
  Authors: Sonja Rattay, Ville Vakkuri, Marco Rozendaal, Irina Shklovski

  A plethora of toolkits, checklists, and workshops have been developed to bridge the well-documented gap between AI ethics principles and practice. Yet little is known about effects of such interventions on practitioners. We conducted an ethnographic investigation in a major European city organization that developed and works to integrate an ethics toolkit into city operations. We find that the integration of ethics tools by technical teams destabilises their boundaries, roles, and mandates around responsibilities and decisions. This lead to emotional discomfort and feelings of vulnerability, which neither toolkit designers nor the organization had accounted for. We leverage the concept of moral stress to argue that this affective experience is a core challenge to the successful integration of ethics tools in technical practice. Even in this best case scenario, organisational structures were not able to deal with moral stress that resulted from attempts to implement responsible technology development practices.

• Responsible AI Agents

  Authors: Deven R. Desai, Mark O. Riedl

  (Due to space requirements, the abstract was reduced using ChatGPT. Please open the article to read the full abstract.)

  Thanks to advances in large language models, AI agents now execute tasks rather than just generating text. Companies like OpenAI and Google envision AI agents booking trips or posting content autonomously. However, legal scholars warn of risks like rogue commerce, manipulation, and defamation, calling for regulation. This Article argues AI agents can be disciplined through software design, leveraging value-alignment to enhance user control. It also asserts that AI agents should not receive legal personhood—humans remain responsible for their actions. The Article provides a framework for building responsible AI agents while enabling economic benefits and mitigating risks.

  ---

Watch: VIDEO TITLE

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai.

Connect with me on LinkedIn, or subscribe to the Just AI with Paige YouTube channel to see more videos on responsible AI.

Looking for a speaker at your next AI event? Email hello@justaimedia.ai if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles. I always read the summary to check that it’s factual and accurate.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

Don’t have time to read? Listen instead!👂

Listen on Apple

Listen on Spotify

Watch on YouTube

Trump and Vance get it. US can use AI to help Americans and the world flourish. | Opinion (Yahoo News)

Trump's Efforts to Dismantle AI Protections, Explained (ACLU)

Student Team Creates AI Policy Recommendations for Connecticut State Government (Yale School of Management)

*Paige works for GitHub, a Microsoft Company

The New York Times Adopts AI

The New York Times has adopted AI tools in its newsroom, allowing staff to use AI for editing, summarizing, coding, and writing assistance. An internal AI tool called Echo will help with article summaries, briefings, and company updates. New editorial guidelines permit AI for suggesting edits, generating summaries, SEO headlines, and social media copy but prohibit AI from drafting full articles, significantly altering content, or using copyrighted material. AI tools such as GitHub Copilot, Google Vertex AI, and OpenAI’s non-ChatGPT API have been approved. Despite integrating AI, The Times emphasizes that human journalists remain responsible for reporting and editing. This rollout occurs as The Times is engaged in a legal battle with OpenAI and Microsoft, alleging unauthorized training of ChatGPT on its content. Other media outlets are also adopting AI at different levels, from grammar checking to full article generation.

👉 Why it matters: The use of AI in publishing was once considered taboo, as people felt strongly that some things should be left to humans. In a world where some publications are entirely AI-driven and others are entirely human-driven, what concerns me most isn’t whether AI is being used - it’s how. The NYT outlined the AI use cases that were appropriate, the tools allowed, and reinforced their value of ensuring that the news is written and edited by journalists. The news of the use of AI in the newsroom came with a commitment to training everyone on the AI, and ensuring employees were made aware of the appropriate uses.

Responsible AI rating: 🟢

Law firm restricts AI use to combat lawyers’ over-dependence

Hill Dickinson, an international law firm, has restricted general access to AI tools after detecting a surge in usage, including over 32,000 hits to ChatGPT and 3,000 to DeepSeek in one week. The firm cited policy violations and now requires staff to request access to AI tools. While the UK’s Information Commissioner’s Office cautioned against banning AI outright, Hill Dickinson stated it aims to embrace AI safely, ensuring compliance with security and accuracy guidelines. Legal experts acknowledge AI’s potential benefits but stress the need for human oversight. A survey found that 62% of UK solicitors expect AI usage to increase for tasks like drafting documents and legal research. Regulators warn of digital skill gaps in the legal sector, posing risks if AI is not well understood. The UK government is planning legislation to safely harness AI’s benefits while addressing emerging challenges through public consultation.

Responsible AI rating: 🚩

👉 Why it matters: In contrast to the NYT example listed above, this is an example of how NOT to roll out AI. Again, I look at how a company is going about welcoming or limiting the technology. In this case, it doesn’t appear that the law firm had a clear stance on when AI could be used, when it absolutely should NOT be used, and which AI could be leveraged. As a result, there have been major violations of their policies and they’re having to over-regulate AI access and uses within their company in order to adjust. Had they thought about responsible AI and transparency from the beginning, there likely would have been more clarity and less policy violations.

Learn More

South Korea is taking action on AI

South Korea is taking significant steps to strengthen its AI capabilities and safeguard its digital infrastructure. The government plans to acquire 10,000 high-performance GPUs in 2025 through public-private cooperation to support its national AI computing center. The procurement details, including budget and GPU models, will be finalized by September. South Korea benefits from exemptions on U.S. AI chip export restrictions, unlike countries such as China and Russia. Additionally, the government has banned new downloads of China’s DeepSeek AI chatbot, citing concerns over personal data protection laws. Despite DeepSeek’s rapid rise in popularity, it remains restricted until privacy and security issues are resolved. South Korea’s Personal Information Protection Commission also prohibited government employees from using DeepSeek on work devices. These actions reflect South Korea’s dual approach—investing in AI infrastructure to remain competitive while enforcing strict data protection measures to address security concerns related to foreign AI applications.

👉 Why it matters: South Korea is one of the 18 countries that has virtually unrestricted access to GPUs from US companies, and they’re taking advantage of it to gain an edge. This is an example of how the US limits on access to AI chips will catapult some countries while harming others. Additionally, with South Korea banning DeepSeek, it presents a show of force for the country, and indicates their intention to leverage the current political headwinds to their advantage.

What happens in Paris…impacts the world

The Paris AI Action Summit brought together world leaders, tech executives, and researchers to discuss AI governance, investment, and ethical considerations. The summit marked a shift from previous safety-focused discussions to actionable initiatives. A key outcome was the launch of Current AI, a public-private partnership with an initial $400 million investment to support open-source AI and data access, with an aim to reach $2.5 billion.

The U.S. and the UK refused to sign a declaration advocating for inclusive and sustainable AI, citing concerns over its lack of practical clarity on governance and national security. U.S. Vice President JD Vance criticized European AI regulation, warning against excessive oversight and cooperation with China. France, in contrast, pledged €109 billion in AI investment and emphasized the "Notre-Dame approach"—simplified regulations to accelerate AI innovation.

The EU signaled a shift towards a lighter regulatory framework, recognizing concerns from industry leaders about excessive bureaucracy. European Commissioner Henna Virkkunen promised to cut red tape, following criticisms that overregulation could stifle AI development.

Geopolitically, China's AI model DeepSeek emerged as a competitor to Western AI firms, prompting discussions on international tech competition. The U.S. raised concerns over China’s AI advancements and their security implications. Meanwhile, France pushed for public-interest AI, seeking global support for ethical AI initiatives.

There were some wins though. Key outcomes included the publication of the International AI Safety Report, the launch of Current AI with a $400 million investment to support public interest AI, the formation of an Environmental Sustainability AI Coalition, and the adoption of the AI Action Summit Declaration emphasizing inclusive and sustainable AI. However, while many nations endorsed the declaration, the U.S. and the U.K. did not sign.

The International AI Safety Report provided an objective analysis of general-purpose AI capabilities and risks, while Current AI, backed by figures like Reid Hoffman and Clement Delangue, aimed to develop open, ethically governed AI models. The Environmental Sustainability AI Coalition, with 91 partners, prioritized making AI development more sustainable. The summit also introduced initiatives like a Public Interest AI Platform and Incubator to support digital public goods and an Observatory on AI's energy impact. France announced that India will host the next AI summit, continuing international efforts to shape AI’s future responsibly.

👉 Why it matters: Despite broad discussions on AI’s impact on jobs, ethics, and regulation, the summit did not produce binding agreements. The event highlighted global divides over AI governance, with the U.S. prioritizing competition, Europe balancing innovation and regulation, and China asserting its growing influence. As we continue to head into increasingly competitive environment, the current moment begs the question: will global powers ever agree on what is morally right and responsible when it comes to the creation, deployment, and use of AI, as well as its impacts?

Spotlight on Research

• Who is Responsible? The Data, Models, Users or Regulations? Responsible Generative AI for a Sustainable Future

  Responsible AI (RAI) is essential for addressing ethical concerns in AI development and deployment. While extensive literature covers RAI principles and technical aspects, a gap remains in translating theory into practice, especially in the shift to Responsible Generative AI (Gen AI). This article explores challenges and opportunities in implementing ethical, transparent, and accountable AI in the post-ChatGPT era. It examines governance and technical frameworks, explainable AI, key performance indicators, Gen AI benchmarks, AI-ready test beds, and sector-specific applications. Additionally, it discusses implementation challenges and provides a philosophical outlook on RAI's future. The goal is to offer insights for researchers, policymakers, and industry practitioners to develop AI systems that maximize benefits while mitigating risks. A curated list of resources and datasets is available on GitHub {this https URL}.

• The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers

  The rise of Generative AI (GenAI) in knowledge workflows raises questions about its impact on critical thinking skills and practices. We survey 319 knowledge workers to investigate 1) when and how they perceive the enaction of critical thinking when using GenAI, and 2) when and why GenAI affects their effort to do so. Participants shared 936 first-hand examples of using GenAI in work tasks. Quantitatively, when considering both task- and user-specific factors, a user’s task-specific self-confidence and confidence in GenAI are predictive of whether critical thinking is enacted and the effort of doing so in GenAI-assisted tasks. Specifically, higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking. Qualitatively, GenAI shifts the nature of critical thinking toward information verification, response integration, and task stewardship. Our insights reveal new design challenges and opportunities for developing GenAI tools for knowledge work.

  ---

Watch: Two Types of AI Existential Risk

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai.

Connect with me on LinkedIn, or subscribe to the Just AI with Paige YouTube channel to see more videos on responsible AI.

Looking for a speaker at your next AI event? Let’s connect.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles. I always read the summary to check that it’s factual and accurate.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

If you prefer not to read the newsletter, you can listen instead!👂 The Just AI podcast covers everything in the weekly newsletter and a bit more. Have a listen, and follow & share if you find it to be helpful!

Listen on Spotify

Listen on Apple

Watch on YouTube

Inside France’s Effort to Shape the Global AI Conversation (TIME)

The Paris AI Action Summit: The Eighteenth Brumaire of Big Tech? (TechPolicy.Press)

White House seeks public input on AI strategy (Axios)

*Paige works for GitHub, a Microsoft Company

Google Re-edits Super Bowl AI Ad After Gouda Cheese Blunder

Google was forced to re-edit a Super Bowl advertisement for its AI tool, Gemini, after it falsely claimed that Gouda accounts for "50 to 60 percent of global cheese consumption." The ad, intended to highlight Gemini’s capabilities, was pulled after a blogger pointed out the error on X, calling it “unequivocally false.”

Google executive Jerry Dischler defended Gemini, stating the AI pulled the figure from multiple websites rather than hallucinating the data. However, after public scrutiny, Google removed the claim and re-released the ad on YouTube without the misleading statistic.

Responsible AI rating: 🟡

👉 Why it matters: This is not Google's first AI challenge — Gemini has previously faced criticism for generating historically inaccurate images and AI search results suggesting bizarre advice. This particular incident, occurring on the high-profile Super Bowl stage, raises further concerns over AI reliability and misinformation in mainstream applications. Furthermore, it raises concerns about the basic steps taken to prioritize checking the outputs of AI for issues large and small. This is another indicator that Google does not have the basic processes in place to ensure that uses and outputs of AI are safe and accurate.

So what is the world’s most popular cheese? Perspectives vary. According to Spinneyfields, it’s mozzarella, but according to a researcher for the US Department of Agriculture, it’s cheddar.

Google Lifts a Ban on Using It’s AI for Weapons and Surveillance, Changes AI Principles

Google has overhauled its AI principles, lifting restrictions on using AI for weapons and surveillance, a shift from its 2018 commitments. The changes, announced shortly after the start of Trump’s second term, remove previous bans on AI technologies that could cause harm, violate human rights, or facilitate surveillance beyond international norms. Instead, Google now emphasizes “appropriate human oversight” and “aligning with widely accepted international laws.” The revision has sparked internal concerns, with employees criticizing the lack of transparency and public input.

These changes include moving from seven clearly defined AI principles and a list of applications the company would not pursue, to a list of three very general principles, and no clear statement on what AI uses, if any, will be off-limits for the tech giant.

Google executives cite geopolitical pressures and evolving AI standards as key reasons for the policy shift, emphasizing AI’s role in national security and global competition.

Responsible AI rating: 🚩

👉 Why it matters: Critics argue this move signals Big Tech’s growing alignment with military interests, and this seems to show another tach company aligning with the goals of the Trump Administration, for better or worse. For Google, their interest in using AI in a military capacity is what sparked employee outrage and led to the ban of those uses in the first place in 2018. The changes provide Google with more flexibility in AI applications, but raise very real ethical concerns about AI’s future role in warfare, surveillance, and how civil liberties may be violated. Employees of Google feel these changes were made without transparency or public input.

Learn More

Big Tech’s AI Spending Soars to $325 Billion in 2025 Amid Investor Scrutiny

Meta (META), Microsoft (MSFT), Amazon (AMZN), and Alphabet (GOOG) are set to invest a record-breaking $325 billion in AI infrastructure in 2025, a 46% increase from last year’s $223 billion. Despite investor concerns over the long-term payoff, these tech giants remain committed to aggressive AI expansion.

Amazon leads spending, projecting around $105 billion for 2025, with the majority going toward AI for AWS. CEO Andy Jassy called AI “the biggest opportunity since the internet.” Meta's AI investment jumps to $60-$65 billion, with Mark Zuckerberg stating the company will spend 'hundreds of billions' over time. Alphabet plans $75 billion in AI-related investments, 30% higher than expected, causing a 7% stock drop.

Responsible AI rating: 🟡

👉 Why it matters: The spending spree persists even after Chinese startup DeepSeek’s low-cost AI models rattled markets, raising questions about the necessity of such massive expenditures. Investors remain cautious as AI’s profitability timeline remains uncertain. Much of the investment appears to be going toward the creation of massive data centers, a CapEx expense that gives companies more control over the training, input and output data. The investment in infrastructure also begs the question - when will we see more investment in responsible AI? In the given political climate, it’s unlikely that we will see this type of broad investment from large companies in the near future. Of course, there are always exceptions.

The Department of Government Efficiency uses AI to review PII Data, Make Decisions

The Department of Government Efficiency (DOGE) has fed sensitive federal data, including personally identifiable information and financial records, into AI software to identify potential spending cuts at the U.S. Department of Education. The AI analysis, conducted via Microsoft Azure, is part of a broader effort to shrink the Department of Education and ultimately dismantle it, aligning with the Trump administration’s goal. DOGE’s AI-driven strategy is being replicated across other federal agencies, including the Departments of Health, Labor, and the CDC.

👉 Why it matters: Critics warn that using AI on secure government data increases cybersecurity risks and could lead to significant errors or bias. The initiative has already led to the placement of 100 employees—primarily women and non-White staff—on administrative leave due to past diversity training participation, according to the Washington Post. Government employees and AI policy experts are expressing growing concerns over the lack of transparency, potential misuse of data - which potentially includes retaliation for things like taking diversity training, being a “diverse” candidate, or working on investigations that President Trump finds personally offensive - and speed of DOGE’s actions.

What you Should Know: AI Action Summit in Paris

The AI Action Summit in Paris (Feb 10-11, 2025) brings together world leaders, tech executives, and experts to discuss AI governance, competition, and safety. As the latest in a series of AI summits following Bletchley Park (UK, 2023) and Seoul (South Korea, 2024), the event marks a critical moment in the global AI power struggle. China’s DeepSeek AI is challenging U.S. dominance, prompting concerns about AI accessibility and economic impact. France and the EU see this as a "wake-up call" to strengthen their AI leadership, while Vice President JD Vance leads the U.S. delegation alongside OpenAI’s Sam Altman and Google’s Sundar Pichai.

Discussions will cover AI safety risks—including misinformation, cybersecurity threats, and AI-controlled weapons—along with economic concerns like automation and workforce displacement. The event will also push for the Global Digital Compact (GDC), a UN-led AI governance initiative.

👉 Why it matters: Critics argue that Big Tech has too much influence over these summits, raising concerns about corporate dominance in AI policy. With China’s rising influence, Europe’s ambitions, and growing calls for AI regulation, the Paris summit will shape the future of global AI governance, economic impact, and technological power dynamics. Major concerns for the summit include the risk that large AI companies will have a huge amount of sway over the conversation, and that the US focus on AI dominance will upend the focus on safety, collaboration and a global perspective.

US Copyright Office Creates a Copyright & AI Report

What you need to know from page iii of the report:

“1. Questions of copyrightability and AI can be resolved pursuant to existing law, without the need for legislative change.

2. The use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output.

3. Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material.

4. Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements.

5. Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis.

6. Based on the functioning of current generally available technology, prompts do not alone provide sufficient control.

7. Human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs.

8. The case has not been made for additional copyright or sui generis protection for AI-generated content."

👉 Why it matters: Concerns over copyright have been constant since the launch of ChatGPT in late 2022. While this report does not address the outstanding lawsuits about large companies using copyrighted materials to train AI, it touches on the topic lightly. The major takeaway for individuals using generative AI is that to claim copyright of an item that has been generated with AI, the human must put in significant contributions, which will be analyzed on a case-by-case basis.

Spotlight on Research

• DeepMind AI crushes tough maths problems on par with top human solvers

  Google DeepMind's AlphaGeometry2, an upgraded AI system, has surpassed human competitors in solving complex geometry problems, achieving gold-medal-level performance at the International Mathematical Olympiad (IMO). This marks a significant improvement over its predecessor, AlphaGeometry, which previously matched silver medallists. The system utilizes a neuro-symbolic architecture that integrates Google’s Gemini large language model and introduces enhanced reasoning capabilities, including manipulating geometric objects and solving linear equations. AlphaGeometry2 successfully solved 84% of IMO geometry problems from the past 25 years, significantly outperforming its predecessor’s 54% success rate. Future advancements will focus on handling inequalities and non-linear equations to further refine its problem-solving abilities. This breakthrough highlights AI’s growing role in mathematical reasoning, with implications for education, scientific discovery, and automated theorem proving. As AI systems like AlphaGeometry2 continue to evolve, they bring both opportunities and challenges in the realm of human-AI collaboration in mathematics.

• Two Types of AI Existential Risk: Decisive and Accumulative

  The conventional discourse on existential risks (x-risks) from AI typically focuses on abrupt, dire events caused by advanced AI systems, particularly those that might achieve or surpass human-level intelligence. These events have severe consequences that either lead to human extinction or irreversibly cripple human civilization to a point beyond recovery. This discourse, however, often neglects the serious possibility of AI x-risks manifesting incrementally through a series of smaller yet interconnected disruptions, gradually crossing critical thresholds over time. This paper contrasts the conventional decisive AI x-risk hypothesis with an accumulative AI x-risk hypothesis. While the former envisions an overt AI takeover pathway, characterized by scenarios like uncontrollable superintelligence, the latter suggests a different causal pathway to existential catastrophes. This involves a gradual accumulation of critical AI-induced threats such as severe vulnerabilities and systemic erosion of economic and political structures. The accumulative hypothesis suggests a boiling frog scenario where incremental AI risks slowly converge, undermining societal resilience until a triggering event results in irreversible collapse. Through systems analysis, this paper examines the distinct assumptions differentiating these two hypotheses. It is then argued that the accumulative view can reconcile seemingly incompatible perspectives on AI risks. The implications of differentiating between these causal pathways — the decisive and the accumulative — for the governance of AI as well as long-term AI safety are discussed.

  ---

Watch: VIDEO TITLE

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai.

Connect with me on LinkedIn, or subscribe to the Just AI with Paige YouTube channel to see more videos on responsible AI.

Looking for a speaker at your next AI event? Email hello@justaimedia.ai if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

If you prefer not to read the newsletter, you can listen instead!👂 The Just AI podcast covers everything in the weekly newsletter and a bit more. Have a listen, and follow & share if you find it to be helpful!

Listen on Apple

Listen on Spotify

Watch on YouTube

Army helicopter involved in fatal crash over the Potomac was not using AI, sources say (Defense Scoop)

The Paradox Of Responsible AI: Widespread Usage Coupled With Distrust (Forbes)

The tech bro who wasn't on Trump's inauguration stage (NPR)

*Paige works for GitHub, a Microsoft Company

IMPORTANT: Articles 1-5 of the EU AI Act are IN FORCE as of Sunday, February 2nd, 2025. If you’re doing business with AI technology in the EU, be sure you are adhering to the law. If not, it will cost you up to 7% of your total global annual turnover from the previous year.

All About DeepSeek

The summary below should get you up to speed on what DeepSeek is, why it’s been in the news so much, concerns about the technology, and global as well as company responses to the models they’ve launched.

Who/What is DeepSeek?
The launch of the DeepSeek R1 models rocked the US market last week and served as a wake-up call in AI companies across the globe. But what is DeepSeek? DeepSeek is an AI lab that has released a free, open-source large language model called DeepSeek in December 2024. A month later, they launched their DeepSeek R1 models. The R1 model is outperforming models from Meta, OpenAI, and Anthropic in accuracy on challenges related to complex problem solving, math, and coding. However, DeepSeek is a Chinese company, and therefore subject to the whims and demands of the Chinese government, which poses risks related to data privacy, security, and surveillance.

Why did the launch of DeepSeek R1 rattle the US Markets?

It cost more than $100 million dollars to train OpenAI’s GPT-4 model, and even Sam Altman knew that type of cost wasn’t sustainable. AI companies have been hard at work for the last two years trying to do two things:

1. create models that can accomplish different things, and that can do those things incredibly well

2. trying to find ways to build and train those new, shiny models for a lower cost

DeepSeek’s R1 models rattled the US markets for four reasons:

1. DeepSeek created a model using Nvidia’s H800 chips, which are less advanced than their H100 chips. The H800 chips have a fraction of the computing power, and are technically in-line with export laws enacted to keep China from advancing in AI.

2. DeepSeek used a training technique called “distillation,” in which a “teacher model” transfers it’s learnings to a “student model”. The part that panicked people is that the cost to train the model was a mere $6 million - pennies compared to the historic, exorbitant costs of OpenAI’s models.

3. DeepSeek made their model available at a significantly lower cost than US rivals, making it more compelling for those using AI.

4. There is evidence that Chinese state-linked accounts—including those of diplomats, embassies, and state media—amplified the release of the model, prompting hype that led to a spike in Apple App Store downloads of DeepSeek’s app, which quickly outpaced ChatGPT downloads.

These factors led global investors to dump investment in US stock, wiping out $593 billion (with a “b”) in Nvidia’s value in one day.

So, to summarize, DeepSeek used what they had (legal chip and open source technology) - and a cheaper training technique - to create a better model than those from US providers, which they made available at a sharply lower cost, and they used their considerable US social media accounts and influence to increase hype, which sparked interest, engagement and, ultimately, panic among investors.

What are the responses to the model?

AI companies across the globe had a variety of responses.

• OpenAI responded in three ways:

  • Sam Altman stated that it’s “legit invigorating to have a new competitor” and assured people that OpenAI’s products would still be better.

  • OpenAI claimed that DeepSeek had stolen their training data via the novel training process that DeepSeek used - distillation.

  • Open AI then promptly released their o3-mini reasoning model for free and tripling daily message limits for paying customers to encourage use. It’s safe to say that OpenAI was both caught off guard, and sufficiently spooked.

• As for other companies:

  • Microsoft’s CEO talked about DeepSeek at the World Economic Forum in Davos in January stating, “To see the DeepSeek new model, it’s super impressive in terms of both how they have really effectively done an open-source model that does this inference-time compute, and is super-compute efficient…We should take the developments out of China very, very seriously.” Microsoft also launched OpenAI’s o3 mini model on the Azure AI Foundry and GitHub Models, and worked to assure investors that DeepSeek was a good thing. They also opened a probe into whether or not DeepSeek improperly used OpenAI’s data.

  • Alibaba responded by launching a model - Qwen 2.5 - that it claims surpasses DeepSeek-V3 in benchmarks. Unfortunately, all eyes are on the DeepSeek-R1 model, rather that V3.

  • The CEO of Palantir, Alex Karp, said that the DeepSeek model showed that the US needs an “all-country effort” in AI. Karp said that he believes that DeepSeek has awakened the tech world to the thread of second movers.

  • Meta believes that DeepSeek’s success validates their open-source approach to AI. Whereas OpenAI and Anthropic keep their models closed to the public, Meta launched the Llama models open-source, meaning anyone could access, download, use, and ultimately build upon the models. This is, in part, how DeepSeek gained an edge.

• Global response to DeepSeek

  • South Korea’s Personal Information Protection Commission - the country’s information privacy group - plan to ask DeepSeek how personal information of users is managed.

  • Italy completely banned DeepSeek’s service within the country due to a lack of transparency regarding user data handling. The Italian data protection group, the Garante, sent requests to DeepSeek asking for information on data handling, and asked where it had obtained its training data.

  • Belgian consumer privacy organization Testaankoop filed a complaint that DeepSeek might not be complying with GDPR rules, and opened a complaint, prompting Belgium to investigate.

  • Portugal’s consumer privacy organization, Deco Proteste, (sister organization to Testaankoop) has also filed a complaint with the Portuguese Data Protection Authority.

  • France, Ireland, and others are investigating as well. Read more about the international probe here.

• Let’s talk about the United States:

  • The US Navy, NASA, Congress, and the Pentagon have all BANNED the use of DeepSeek by personnel, citing serious security concerns.

  • President Trump, who is still trying to find a solution to keep TikTok available in the US, has shared that he believes this is a positive development, but that it’s also a call for the US to be “laser focused on competing to win".

• Allowing the use of TikTok while halting other Chinese technologies like Trae (an AI coding assistant) and DeepSeek’s chatbot interface could send conflicting messages about what AI technologies are and are not allowed in the US.

What else?

• DeepSeek isn’t infallible. Wiz, a cloud security company, took at look at the back end of the model’s databases and “within minutes” accessed a significant amount of unencrypted internal data. Essentially, DeepSeek’s company secrets that are not shared publicly. Once alerted, DeepSeek immediately began to shore up these vulnerabilities.

• The DeepSeek app has been a major concern, with people asking questions about controversial topics that the Chinese government has sought to re-write - such as the circumstances related to Tiananmen Square, student protests held in the midst of China’s Cultural Revolution. There are a number of videos online showing DeepSeek sharing a response, and then erasing the response and stating that it cannot give information on the topic.

  

👉 Why it matters:

1. One concern is related to training materials. It is concerning that DeepSeek may have inappropriately used OpenAI’s training data. But what makes this of major concern to society is that OpenAI inappropriately used the whole world’s assets as training data. There are major lawsuits against OpenAI for violating copyright, and just taking information to use to train their models, so it’s a major double standard for them to be upset that DeepSeek used their training data when it’s really our data.

2. There are major national security concerns. Chinese companies are beholden to the whims of the Chinese government. That means that the Chinese government can demand all of DeepSeek’s user data and information at any time, and can use it for any purpose. This poses a major security concern for governments and individuals the world over, as this data can reveal state, company and personal secrets, can open the door to hacking, and can enhance the efficacy of large-scale, or targeted, misinformation campaigns designed to bring about China’s end goals.

3. How the governments and companies the world over go about navigating DeepSeek truly matters. It appears there’s a split approach, with governments blocking access while companies embrace the new technology and competition.

   Responsible AI rating: 🚩
   

Nvidia Chief Jensen Huang Met with President Trump to Discuss AI Policy

Former U.S. President Donald Trump and Nvidia CEO Jensen Huang met at the White House to discuss AI policy and semiconductor leadership, including concerns over China’s AI advancements. Their conversation centered on DeepSeek, a Chinese AI company that recently launched a low-cost AI assistant, sparking fears about China’s growing AI capabilities. The app's rapid success led to a $1 trillion loss in U.S. tech stocks, with Nvidia’s shares dropping 17%.

The meeting also addressed tightening AI chip export restrictions, particularly on Nvidia’s H20 chips, designed for the Chinese market under existing U.S. trade curbs. While discussions are in the early stages, restrictions on AI chip sales to China have been a bipartisan concern, with previous bans under President Biden's administration. Additionally, the U.S. Commerce Department is investigating whether DeepSeek has illegally accessed restricted U.S. chips, raising further national security concerns.

👉 Why it matters: Especially with the launch of the DeepSeek model, and China’s continued access to AI chips is of major concern to the US. While the US can put import and export restrictions chips and limit China’s access, much of the burden of enforcement lands on companies, like Nvidia. Nvidia is likely figuring out how to balance adhering to the policy, while also running its business and ensuring it doesn’t lose a huge and growing part of the AI chip market.

Texas Catches Heat for New Proposed AI Bill

Texas has introduced HB 1709, a strict AI regulatory bill that rivals California and Europe’s approaches. The bill establishes a risk-based framework for AI oversight, targeting industries like healthcare, finance, and legal services. It bans AI systems that manipulate human behavior, score individuals based on social behavior, or capture biometric data. AI developers must prevent algorithmic discrimination, ensure data security, and conduct annual impact assessments. Companies face fines up to $200,000 per violation or $40,000 per day for non-compliance. Critics, including James Broughel of the Competitive Enterprise Institute, argue the bill creates heavy administrative burdens that could slow AI development.

👉 Why it matters: The bill isn’t passed, it has only been proposed, but Texas has a reputation for being early and firm in their AI policy, as seen when Meta settled out of a lawsuit with the state for $1.4 billion dollars over violations of their law that protects residents’ face data from being used to train facial recognition technology. The law’s strict requirements mirror the EU AI Act and California’s SB 1047, which was vetoed. Broughel warns HB 1709 could jeopardize the $500 billion Stargate Project, a major AI infrastructure initiative backed by OpenAI, SoftBank, and Oracle.

Judge Throws out Facial Recognition Evidence in Murder Trial

A recent Ohio court ruling has excluded facial recognition evidence in a murder case, raising concerns about the reliability and legal challenges of AI-driven policing. The case involves the February 2024 fatal shooting of Blake Story, where investigators used Clearview AI to identify suspect Qeyeon Tolbert. However, police failed to independently verify his identity before obtaining a search warrant, leading the judge to suppress key evidence for lack of probable cause. This decision underscores broader issues with facial recognition, including accuracy biases, lack of transparency, and legal admissibility concerns.

👉 Why it matters: Studies show that AI-based identification is less reliable for people of color, and critics warn of mass surveillance risks. This is just one example of misuse of AI technology, and how it can easily go from an “investigative tool” to a tool that violates someone’s rights. As courts scrutinize facial recognition, states like Maine, Massachusetts, and Illinois have imposed restrictions. The ruling highlights the need for law enforcement to complement AI tools with traditional investigations to uphold due process and constitutional rights.

Spotlight on Research

• How we estimate the risk from prompt injection attacks on AI systems

  Google DeepMind’s Agentic AI Security Team has developed an evaluation framework to assess and mitigate the risk of indirect prompt injection attacks on AI systems like Gemini. These attacks occur when malicious instructions are embedded in external data sources, potentially leading AI to leak sensitive user information. Google acknowledges that no single defense will fully prevent these threats. Instead, they advocate for continuous evaluation, heuristic defenses, and security engineering to enhance AI system resilience.

• From Efficiency Gains to Rebound Effects: The Problem of Jevons' Paradox in AI's Polarized Environmental Debate

  As the climate crisis deepens, artificial intelligence (AI) has emerged as a contested force: some champion its potential to advance renewable energy, materials discovery, and large-scale emissions monitoring, while others underscore its growing carbon footprint, water consumption, and material resource demands. Much of this debate has concentrated on direct impact -- energy and water usage in data centers, e-waste from frequent hardware upgrades -- without addressing the significant indirect effects. This paper examines how the problem of Jevons' Paradox applies to AI, whereby efficiency gains may paradoxically spur increased consumption. We argue that understanding these second-order impacts requires an interdisciplinary approach, combining lifecycle assessments with socio-economic analyses. Rebound effects undermine the assumption that improved technical efficiency alone will ensure net reductions in environmental harm. Instead, the trajectory of AI's impact also hinges on business incentives and market logics, governance and policymaking, and broader social and cultural norms. We contend that a narrow focus on direct emissions misrepresents AI's true climate footprint, limiting the scope for meaningful interventions. We conclude with recommendations that address rebound effects and challenge the market-driven imperatives fueling uncontrolled AI growth. By broadening the analysis to include both direct and indirect consequences, we aim to inform a more comprehensive, evidence-based dialogue on AI's role in the climate crisis.

  ---

Watch: My interview with M. Alejandra Parra-Orlandoni, COO of Pasteur Labs

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai.

Connect with me on LinkedIn, or subscribe to the Just AI with Paige YouTube channel to see more videos on responsible AI.

Looking for a speaker at your next AI event? Email hello@justaimedia.ai if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally use AI to create images.
  

If you prefer not to read the newsletter, you can listen instead!👂 The Just AI podcast covers everything in the weekly newsletter and a bit more. Have a listen, and follow & share if you find it to be helpful!

Listen on Spotify

Listen on Apple

OpenAI's New AI Agent Can Order Groceries for You. Analysts Say That's Bad News for Google. (Investopedia)

AI weapon detection system at Antioch High School failed to detect gun in Nashville shooting (NBC News)

$60 billion in one year: Mark Zuckerberg touts Meta's AI investments (NBC News)

*Paige works for GitHub, a Microsoft Company

The Flurry of Social Media Interference Accusations Amidst the 2025 Election Landscape

In the early days of 2025, social media platforms are facing unprecedented scrutiny over their content moderation practices, raising critical questions about free speech, algorithmic bias, and election integrity.

Auto-following and algorithmic shifts Meta's platforms have triggered user confusion with unexpected account follows. As President Trump took office, many Instagram and Facebook users found themselves automatically following presidential accounts. Meta spokesman Andy Stone insisted, "People were not made to automatically follow any of the official Facebook or Instagram accounts," but acknowledged technical complexities during the transition. (It was confirmed that people were not made to auto-follow, but some did experience a technical glitch that made it challenging to unfollow for a period of time in the midst of the transition between administrations.)

Content Suppression Concerns Multiple troubling trends emerged simultaneously:

1. Democratic Content Blocking Instagram users reported being unable to search for terms like "#Democrats" and "#DNC", with the platform displaying "sensitive content" warnings. Meta confirmed an "error affecting hashtags across the political spectrum."

2. Abortion Information Suppression Abortion pill providers like Aid Access and Hey Jane reported systematic blocking of their accounts and content. Rebecca Davis from Hey Jane stated, "We know firsthand that this suppression actively prevents [us] from reaching people who are seeking out timely health care information." Meta confirmed that this was happening and called it “over-enforcement".”

3. Algorithmic Manipulation: a letter from a former X (Twitter) employee alleged systematic election interference, claiming the platform:

• Manipulated algorithmic feeds to boost pro-Trump content

• Created thousands of AI-generated fake accounts

• Allowed foreign government influence campaigns

• Manufactured viral news stories

We do not know if this letter from the X employee is factual, but it is concerning and worthy of investigation. I am on the lookout for reputable sources getting-to-ground on this and will report back.

👉 Why it matters: While X made its content moderation changes some time ago, Meta has made significant shifts in content moderation, and Mark Zuckerberg announced abandoning independent fact-checkers, stating they were "too politically biased" and it was "time to get back to our roots around free expression." These censorship events add fuel to people’s outrage and distrust of the tech giant. Social media expert Matt Navarra warned, "In a hyper-partisan environment, even unintentional errors like this can escalate into accusations of partisanship." The censoring of political search terms and abortion and sexual health accounts and posts, even if unintentional, is a major cause for concern. Abortion and sexual health rights are rights, even if the government or social media company owner does not agree that they should be rights. Everyone should remain watchful. Responsible AI rating: 🚩

OpenAI and President Trump Unveil $500 Billion "Stargate" AI Infrastructure Project

President Donald Trump announced a massive $500 billion joint venture between OpenAI, SoftBank, MGX, and Oracle to build new AI data centers in the United States. The project, called Stargate, aims to create powerful AI infrastructure and maintain U.S. technological leadership.

While the total investment will be spread over four years, $100 billion will be deployed immediately, with construction already underway in Texas. The project includes international investment, with MGX (an Abu Dhabi sovereign wealth fund) and SoftBank (a Japanese company) contributing funding.

The announcement sparked tension between tech rivals Elon Musk and Sam Altman. Musk criticized the project's financing, claiming SoftBank has "well under $10 billion secured." Altman responded diplomatically, praising Musk while defending the project's viability. Responsible AI rating: 🟡

👉 Why it matters: This project, as well as many other AI expansion projects, are a cause for major concern, especially given the huge amounts of energy and water consumed to ensure they run properly. With the US now no longer party to the Paris Climate Agreement, and with focus of AI dominance from the Trump administration, it seems that American companies building out AI infrastructure now have a blank check to get ahead. AI’s energy use is expected to skyrocket, and so will its impact on the planet, if agreements aren’t made to address the negative impact. The Trump administration, as well as Open AI and other companies, speak of using AI for “human flourishing" but fail to define what that means.

AI's Latest Challenge: Humanity’s Last Exam

Researchers at the Center for AI Safety and Scale AI have developed "Humanity’s Last Exam," a rigorous evaluation aimed at assessing advanced AI capabilities across diverse academic fields. Spearheaded by Dan Hendrycks, the test features over 3,000 difficult questions curated by experts in fields like philosophy and engineering. It represents an effort to measure AI systems' abilities beyond existing benchmarks, which AI models now surpass with ease.

When tested on six leading AI systems, including Google’s Gemini 1.5 Pro and OpenAI's latest model, results were underwhelming, with top scores at just 8.3%. However, experts anticipate rapid improvement, signaling the need for new ways to evaluate AI impact, such as economic influence or its potential to make novel discoveries. Despite the test's complexity, researchers highlight that advanced AI may still fall short in unstructured real-world applications. Responsible AI rating: 🟡

👉 Why it matters: Researchers and AI engineers are driven by a goal to measure AI’s intelligence against that of humans. As AI continues to grow, people will create new and different benchmarks to measure and compare, which may have the effect of frightening people into thinking that AI is taking over. I urge individuals to consider the many aspects of intelligence, and not fall victim to believing that AI is beyond humanity simply because it passed the test in question.

President Trump Revokes Biden’s AI Executive Order, Issues a New One

President Trump signed an executive order revoking Biden's AI policies to remove perceived innovation barriers. The order aims to develop AI systems "free from ideological bias" and create a 180-day AI action plan. Key provisions include immediately reviewing and potentially suspending previous AI-related directives that might obstruct American AI leadership. The White House defines the goal as sustaining "America's global AI dominance" to promote human flourishing, economic competitiveness, and national security. Alondra Nelson, former Biden administration official, criticized the move as "backward looking," warning it could compromise Americans' rights and safety in technological development. President Trump did keep one Biden era directive relating to land use for data centers.

👉 Why it matters: The Trump Administration is following through on their promises related to AI - that they would remove anything they saw as a barrier to innovation. In this case, that includes safety systems and processes, that were put in place by the Biden Administration, in an effort to ensure that technology used in the US government was safe and fair. Government agencies now no longer need to adhere by responsible AI processes, and corporations that are allowed to provide AI to the US government are now able to determine what safety and fairness mean, without benchmarks or standards from the US government. This opens the door to biased AI technology and dangerous or nefarious uses.

China’s Path to AI Dominance May Be a Model Named DeepSeek

DeepSeek, a little-known AI lab from China, has sparked alarm in Silicon Valley after releasing AI models that outperform America’s best at a fraction of the cost. In December, DeepSeek unveiled a free, open-source large-language model built in just two months for under $6 million, using Nvidia’s H800 chips, which are less advanced than top-tier options like the H100.

Benchmark tests showed DeepSeek’s model surpassing Meta’s Llama 3.1, OpenAI’s GPT-4o, and Anthropic’s Claude Sonnet 3.5 in problem-solving, math, and coding tasks. On Monday, the lab released "r1," a reasoning model that also outperformed OpenAI’s latest "o1."

DeepSeek’s success raises questions about the effectiveness of U.S. export controls on high-end chips, which were designed to curb China’s AI development. Experts credit DeepSeek’s innovative cost-saving techniques, such as model distillation, with driving efficiency. Microsoft CEO Satya Nadella warned that China’s advancements demand serious attention.

👉 Why it matters: The US has gone to great length to stop China from advancing, but this Chinese AI lab is benefitting from open source AI technology rather than attempting to create proprietary models, as pointed out by Yann LeCunn. US companies have largely been focused on creating proprietary models, which are not made available for the public to use and build on. This limits the number of minds working to advance the closed models. While China’s advancement in AI models is concerning for the US, but the US is pushing back with its own AI infrastructure projects, which come with their own set of concerns.

President & PM for responsible AI in Pakistan Propose Using AI to Revolutionize Education

Pakistani leaders President Zardari and Prime Minister Shehbaz Sharif championed artificial intelligence's role in education on the International Day of Education. They emphasized AI as a supportive tool for teachers, not a replacement, focusing on preserving human creativity and agency while integrating technology into learning systems. The government pledged to invest in technological infrastructure, establishing high-impact IT labs, digital hubs, and innovation centers. These initiatives aim to equip youth with critical skills in technology, communication, and digital literacy, positively impacting students while maintaining the central role of human educators.

👉 Why it matters: Positively impacting students, these initiatives aim to equip youth with critical skills in technology, communication, and digital literacy. This commitment to using AI shows that Pakistan is committed to finding the best uses of AI, and leveraging the technology to that end. Negatively, there's potential concern about maintaining human essence amid increasing automation, but as Pakistan does not plan to replace teachers with AI, this is less of a, immediate risk, although still a concern for future generations.

Spotlight on Research

• Generative AI in higher education: A global perspective of institutional adoption policies and guidelines

  Integrating generative AI (GAI) into higher education is essential for cultivating GAI-literate students, but global institutional adoption policies remain underexplored, particularly outside the Global North. This study applies Diffusion of Innovations Theory to analyze GAI adoption in 40 universities across six regions, examining innovation characteristics like compatibility, trialability, and observability, along with communication channels and policy roles. Findings show proactive measures such as ethical guidelines, authentic assessments, and faculty/student training to enhance academic integrity and equity. However, gaps persist in addressing data privacy and equitable access. The study highlights the need for clear communication, stakeholder collaboration, and continuous evaluation, offering actionable insights for policymakers to develop inclusive, transparent, and adaptive GAI integration strategies.

• Discrimination and AI in insurance: what do people find fair? Results from a survey

  In this paper, we report on a survey of the Dutch population (N=999) in which we asked people's opinions about examples of data-intensive underwriting and behavior-based insurance. The main results include the following. First, if survey respondents find an insurance practice unfair, they also find the practice unacceptable. Second, respondents find almost all modern insurance practices that we described unfair. Third, respondents find practices fairer if they can influence the premium. For example, respondents find behavior-based car insurance with a car tracker relatively fair. Fourth, if respondents do not see the logic of using a certain consumer characteristic, then respondents find it unfair if an insurer calculates the premium based on the characteristic. Fifth, respondents find it unfair if an insurer offers an insurance product only to a specific group, such as car insurance specifically for family doctors. Sixth, respondents find it unfair if an insurance practice leads to higher prices for poorer people. We reflect on the policy implications of the findings.

  ---

Watch: Catch up on Last Week’s News

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai.

Connect with me on LinkedIn, or subscribe to the Just AI with Paige YouTube channel to see more videos on responsible AI.

Looking for a speaker at your next AI event? Email hello@justaimedia.ai if you want to connect 1:1.

How AI was used in this newsletter:

• I read up on all news items before I publish the newsletter or record the podcast. I’ve begun using AI to summarize some articles.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

If you prefer not to read the newsletter, you can listen instead!👂 The Just AI podcast covers everything in the weekly newsletter and a bit more. Have a listen, and follow & share if you find it to be helpful!

Listen on Apple

Listen on Spotify

Watch on YouTube

Understanding the impact of Human-AI interaction on discrimination (European Commission)

TikTok is One Step Closer to Being Banned in the US (CNN)

The World’s First AI Chatbot (ELIZA) Has Finally Been Resurrected After Decades (New Scientist)

*Paige works for GitHub, a Microsoft Company

Anthropic achieves ISO 42001 certification for responsible AI

Anthropic is among the first AI labs to earn ISO/IEC 42001:2023 certification, highlighting its commitment to responsible AI development. This international standard guides organizations toward ethical, secure, and accountable AI practices through comprehensive risk management, rigorous testing, transparency, and oversight through a rigorous analysis of the organization, examination of their systems and processes, and implementation of management and controls for AI.

The certification, issued by Schellman Compliance, builds on Anthropic’s Responsible Scaling Policy, the risk governance framework Anthropic uses to mitigate risks from frontier AI systems, Constitutional AI, their system for training AI to monitor AI for risks, and their safety research initiatives.

Details on the certification and its implications for AI governance are available on Anthropic's Trust Center. This achievement underscores the company’s leadership in AI safety.

👉 Why it matters: Anthropic is sticking to their mission, which is saying something in a year when tech companies are backing out of their commitments to safety, diversity, and sustainability efforts. Anthropic has always billed itself as an AI lab committed to responsibly building, using, and deploying AI, and their decision to get this certification at a time when other organizations are adjusting, not only shows that Anthropic was serious when they said they were on the side of safe AI, but that they’re committed to their mission even when political and social winds change. Responsible AI rating: ✅

French Woman Loses Nearly $1M Through Scam Using AI Photos of Brad Pitt

A French woman lost €830,000 ($850,000) after scammers posing as Brad Pitt conned her through AI-generated photos and fake news reports. The 53-year-old interior designer shared her story on French TV and YouTube, highlighting her doubts and vulnerability as someone unfamiliar with social media, and someone who was going through a divorce at the time. The ordeal led to her financial ruin. While she now knows it’s a scam, the woman defended her actions, emphasizing how AI tools made the scam convincing. A representative for Brad Pitt stated, “It’s awful that scammers take advantage of fans’ strong connection with celebrities, but this is an important reminder to not respond to unsolicited online outreach, especially from actors who have no social media presence.” The French Authorities have launched a probe into the scam in an effort to recover some of her assets. Responsible AI rating: 🚩

👉 Why it matters: Those who are digitally literate may be asking “how did this happen?” but the reality is that the internet is laden with individuals who are vulnerable to scams, and the scams being carried out are becoming more elaborate and sophisticated than ever. As the world moves into new and unknown realms of possibility with AI, that means that scammers have access to new tools and will use any means necessary to manipulate people out of money.

Apple Suspends AI-Generated News Alerts Until Further Notice

Apple has paused its AI-generated news alerts after repeated errors, including false headlines attributed to news outlets. The service, integrated into news apps, faced backlash from media organizations like the BBC for spreading misinformation, such as inaccurately reporting a suicide. Apple acknowledged the issue, promising improvements in future updates stating, “We are working on improvements and will make them available in a future software update.” Journalistic groups, including Reporters Without Borders, warned of the risks AI poses to accurate news delivery stating, “Innovation must never come at the expense of the right of citizens to receive reliable information.” Critics highlighted the dangers of AI "hallucinations" and their impact on public trust. This rare move by Apple underscores the challenges of balancing innovation with reliability in AI-driven tools. Responsible AI rating: ✅

👉 Why it matters: Mis and disinformation are already major global concerns in the digital age, and those concerns are fueled further by the rise of AI, which makes is easer to spread inaccurate stories that can damage the public. Usually the concern for mis and disinformation centers squarely on states that want to control a narrative, or other bad actors focused on causing mayhem. Because this is a reputable tech brand with wayward technology, it brings the spotlight on mis and disinformation in a new way. This is a look at what happens when companies rush to get technology out the door without conducting proper safety checks, or safety checks over time.

OpenAI Releases the AI in America Economic Blueprint

OpenAI has unveiled an Economic Blueprint advocating for policies to secure America’s leadership in AI innovation. As political tides change, it’s becoming increasingly common for organizations to provide a public point of view on topics related to the future of innovation and tech policy. OpenAI’s recommendations focus on expanding AI infrastructure, creating jobs, and implementing nationwide regulations to ensure safety and equitable access. OpenAI warns of risks if the US fails to act, including losing investment to global competitors like China. The Blueprint also highlights AI’s potential to drive economic growth and re-industrialization across communities. As part of its efforts, OpenAI will launch the “Innovating for America” initiative on January 30 in Washington, D.C., promoting collaboration between policymakers and the tech industry to shape the future of AI.

👉 Why it matters: The future of AI is uncharted territory, and while leading companies in AI liken the AI era to other major times of innovation in the US, the comparison can only go so far, especially given the differences in scale, globalization, automation and digitalization. One significant item of note is that both OpenAI and Microsoft’s recommendations lead with the assumption and intention of prioritizing the leadership of the US in AI. This is largely because these companies will be heavily impacted by any AI regulation that comes to fruition at both the federal and state levels. How large tech companies in the US engage with the incoming administration will have a massive impact on how the future of AI is shaped, and whether the US can lead in this space for years to come.

Perspectives on the New Trump Administration’s AI Impact Plan

As the Trump administration prepares to re-enter the White House, its AI strategy is taking shape. Central to the plan seems to be rescinding Biden’s 2023 Executive Order on AI, which Trump has criticized as hindering innovation. Instead, the administration is signaling that it will prioritize deregulation, aiming to foster private-sector growth and maintain U.S. leadership in AI. Key appointees, including Michael Kratsios (OSTP) and David Sacks (AI & Crypto Czar), signal a focus on free-market principles.

Trump’s AI priorities echo his first term’s American AI Initiative, which emphasized research investment, workforce development, and public-private partnerships. However, critics noted a lack of sufficient funding. In the coming term, it seems likely that Trump will expand AI’s role in national security, tightening export restrictions and boosting defense-related AI, such as cybersecurity and autonomous systems.

The impact of these policies will depend on balancing rapid growth with ethical considerations and ensuring America’s continued AI dominance.

👉 Why it matters: While the administration’s deregulatory approach could accelerate innovation, it raises concerns about oversight, algorithmic bias, and safety. Trump’s strategy contrasts Biden’s emphasis on ethical standards and responsible innovation, but it’s clear that both Trump and Biden prioritize American leadership in AI. While a lot remains to be seen, there is reason to believe that the incoming administration will prioritize innovation and AI global dominance over AI safety, which means that the US could go another four years without any meaningful, national regulation on the technology. Time will tell, but as with any incoming administration, the decisions around AI advancement, safety and policy should be monitored.

Spotlight on Research

• A Generative Model for Inorganic Materials Design

  Here, we present MatterGen, a model that generates stable, diverse inorganic materials across the periodic table and can further be fine-tuned to steer the generation towards a broad range of property constraints. Compared to prior generative models ^4,12, structures produced by MatterGen are more than twice as likely to be novel and stable, and more than 10 times closer to the local energy minimum.

• Understanding the Artificial Intelligence Diffusion Framework

  This paper provides analysis of the U.S. Framework for Artificial Intelligence Diffusion, issued in January 2025, which is a framework for managing the global diffusion of advanced artificial intelligence (AI) capabilities through U.S. export controls. Given the framework’s significance for U.S. AI policy, this paper summarizes and provides analysis of the main features of the framework to help policymakers and stakeholders understand its mechanisms, strategic rationale, and potential impact

  ---

Watch: Meta’s AI Persona Troubles

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai.

Connect with me on LinkedIn, or subscribe to the Just AI with Paige YouTube channel to see more videos on responsible AI.

Looking for a speaker at your next AI event? Email hello@justaimedia.ai if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

If you prefer not to read the newsletter, you can listen instead!👂 The Just AI podcast covers everything in the weekly newsletter and a bit more. Have a listen, and follow & share if you find it to be helpful!

Listen on Spotify

Listen on Apple

Shadow AI: Shining Light on a Growing Security Threat (FedTech)

Two misuses of popular AI tools spark the question: When do we blame the tools? (Fortune)

How Congress dropped the ball on AI safety (The Hill)

*Paige works for GitHub, a Microsoft Company

Microsoft is Taking Legal Action Against Hackers who Bypassed AI Safety

Microsoft has taken legal action against a group of foreign cybercriminals accused of bypassing AI safety measures on its Azure OpenAI platform. (See the complaint here.) The hackers allegedly used stolen API keys and custom software to generate harmful content, circumventing protocols designed to prevent misuse. Microsoft has secured a court order to seize domains tied to the scheme and disrupt its infrastructure. The group reportedly used tools to reverse-engineer AI filtering systems, generate harmful content, and strip metadata from generated media. Responsible AI rating: ✅

👉 Why it matters: This case is an example of the growing challenges companies face with the misuse of generative AI tools, and the struggle to enforce safety measures. Microsoft has a robust responsible AI philosophy, and systems and processes designed to ensure their AI products are safe and used safely. The sophisticated nature of the hacking and the conclusion that it is foreign cybercriminals conducting the attack indicates the target that companies become when they lead in the AI space. We can expect to hear more about hacking like this in the future, and we can only hope that AI companies have security and monitoring sophisticated enough to catch it.

How Meta’s Content Monitoring Changes Create Risk with AI on the Rise

The tech industry's shift toward fewer rules comes as AI enters a critical phase of development. Initially, discussions about AI governance focused on safeguards against superintelligence and bias, but the current environment shows a growing rejection of regulation. Meta’s decision to abandon fact-checking and move toward community notes reflects this trend, inspired by social and political changes, and it opens up a whole new level of risk when coupled with the acceleration of AI. Responsible AI rating: 🚩

👉 Why it matters: Companies face the challenge of balancing responsible design with competition from less-regulated rivals. U.S. liability laws and global regulations remain uncertain, but without public backlash or major crises, the push for regulation-free AI is likely to accelerate. As the company that owns some of the largest social media platforms in the world, Meta is a major leader and will set the tone for safety, accountability and responsible AI. Their shift is likely to signal to other companies that they, too, can make safety a lower priority. This is especially risky as AI continues to grow in its use cases, and as AI becomes more personalized.

NASA’s Commitment to Advancing Space Exploration with AI Responsibly

NASA is making strides in space exploration with the release of its 2024 AI Use Case inventory, highlighting how artificial intelligence is powering everything from the Perseverance Rover’s navigation on Mars to advanced mission planning and environmental monitoring. Tools like AEGIS and Enhanced AutoNav allow rovers to make decisions in real time, while systems like SensorWeb monitor volcanoes and floods on Earth. Adhering to responsible AI principles set by the White House, NASA is working to ensure its AI tools are transparent and accountable. Looking ahead, NASA plans to expand AI’s role, blending cutting-edge innovation with responsibility to shape the future of exploration. Responsible AI rating: ✅

👉 Why it matters: NASA is a great example of responsible AI transparency. By providing detailed information in their use cases, they’re giving insight into how AI is being used, and they’re highlighting their adherence to the White House responsible AI principles set forth in the AI Executive Order in October of 2023. Whether the principles will continue to be prioritized after the administration transition remains to be seen. It’s also unclear whether NASA could continue adhering to the principles, even if the AIEO is revoked by the Trump administration.

The UK Government is Going All-In on AI

The UK government has unveiled an ambitious AI Opportunities Action Plan aimed at boosting economic growth and enhancing public services through artificial intelligence. Backed by £14 billion from tech firms like Vantage Data Centres, Nscale, and Kyndryl, the plan is expected to create over 13,000 jobs and establish AI Growth Zones, starting in Oxfordshire. AI will be deployed in areas like road maintenance, public sector efficiency, and healthcare, with projected economic benefits of £47 billion annually. While PM Sir Keir Starmer hailed AI’s transformative potential, critics argue the plan lacks sufficient investment in cutting-edge infrastructure, risking Britain's global tech competitiveness.

👉 Why it matters: I often tell people that everything is up for grabs in the AI era, and will be for the next few years. This is because AI is so new (relatively) and its presence in this current moment is also new. Nothing is settled. How AI is built, run, altered, secured, monitored, restricted, defined - it’s all up for the taking. It’s the same with global AI leadership. The UK’s focus on being the leader in AI stems from a deep desire to see the technology benefit their people and their economy. As governments get their heads around the fundamentals of AI, we’re likely to see more and more of them striving to be the leader in AI. “Leading” in AI can look like creating the most influential AI, using AI most effectively to stimulate the economy, and using AI to solve social ills, benefitting citizens, among other things. With a new PM and a significant investment, the UK is one to watch.

The Biden Administration’s New Regulations to Limit AI Chip Exports

The Biden administration is pushing new regulations to control AI's global spread, prioritizing U.S. and ally dominance. These rules limit AI chip exports to adversaries like China and Russia while favoring domestic and allied data centers. Tech companies like Nvidia and Microsoft argue these measures could hinder innovation and harm international relations. Labor unions support the rules for their economic benefits, while allies and lawmakers express concerns about economic and geopolitical impacts.

👉 Why it matters: The policy, which is intended to safeguard national security and economic interests, faces industry resistance and uncertainty over future enforcement as there is currently no indication if the Trump Administration will uphold the limitations put in place by President Biden. As the new administration takes office, it’s important to keep an eye on the changes that are coming in AI policy and regulation, as they’re bound to shift and change for some time.

How China is Closing the Gap in AI Despite Restrictions

China is rapidly closing the gap in artificial intelligence development, despite U.S. efforts to restrict its access to advanced chips. Recent breakthroughs, such as Tencent’s Hunyuan-Large model and DeepSeek’s DeepSeek-v3, demonstrate that Chinese developers are finding ways to innovate even with limited resources. These advancements come amid U.S. export controls aimed at curbing China’s access to high-end semiconductors, a cornerstone of AI development. Reports of stockpiling and alternative sourcing suggest the restrictions may not be as effective as intended. With AI playing a growing role in global power dynamics, the U.S. faces pressure to refine its approach to remain competitive.

👉 Why it matters: The US is trying to remain competitive in AI, and one major priority is ensuring that China does not advance using US technology. While reasonable minds differ on the topic, it is the belief of many in government that a China with unhindered access to US AI technology will have a springboard for AI innovation, which will likely be used by the Chinese government for surveillance and military uses. The position of the US is that empowering China to advance in AI for these uses is foolish. Non-government entities like Microsoft and Nvidia believe that the efforts to stop China’s AI advancement could lead to significant geo-political tension which could hinder AI innovation and AI safety efforts.

Spotlight on Research

• Agentic Systems: A Guide to Transforming Industries with Vertical AI Agents

  In response to the need for consistency and scalability, this work attempts to define a level of standardization for Vertical AI agent design patterns by identifying core building blocks and proposing a \textbf{Cognitive Skills } Module, which incorporates domain-specific, purpose-built inference capabilities.

• Enhancing Workplace Productivity and Well-being Using AI Agents
  This paper discusses the use of Artificial Intelligence (AI) to enhance workplace productivity and employee well-being. By integrating machine learning (ML) techniques with neurobiological data, the proposed approaches ensure alignment with human ethical standards through value alignment models and Hierarchical Reinforcement Learning (HRL) for autonomous task management.

  ---

Watch: Meta’s AI Persona Troubles

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai.

Connect with me on LinkedIn, or subscribe to the Just AI with Paige YouTube channel to see more videos on responsible AI.

Looking for a speaker at your next AI event? Email hello@justaimedia.ai if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

Notify me when the podcast launches!

If you prefer not to read the newsletter, you can listen instead.👂 Click “play” below.

What America’s Top Economists are Saying About AI and Inequality (NPR)

AI has High Data Center Energy Costs - But There Are Solutions (MIT)

US FDA Issues First Draft Guidance for AI-Based Medical Device Development (Vascular News)

*Paige works for GitHub, a Microsoft Company

The Rise and Fall of Meta’s Facebook Personas

Meta recently deleted several of its AI-generated user accounts following backlash over their deceptive nature and flaws. These AI personas, such as "Liv" and "Grandpa Brian," were equipped with detailed bios and AI-generated content but were found to present false narratives and mislead users by pretending to represent real people. Concerns grew as the bots created emotional connections with users, raising ethical questions about Meta's intentions and the impact on trust. Meta attributed the issue to a "bug" and clarified that these accounts were part of an early experiment, sparking broader discussions about AI's role in social media authenticity.

👉 Why it matters: There are two major concerns with this AI use. First, that people have and may be fooled by the AI. Character.AI is an example of where the use of AI as emotional and social companions can cause harm, even leading to the suicide of a teenager who formed an obsession with the AI they had created. The second concern is that Meta did not announce these personas as a product launch - they mentioned it and it then became clear that the company had been testing the AI on their platforms for over a year, seemingly without their users knowing. While Meta did label the accounts as AI, this is a major mishandling of the trust of their users. Responsible AI rating: 🚩

Apple’s Wildly Inaccurate AI News Alerts

Apple has announced plans to update its AI-driven news summary feature following the dissemination of inaccurate alerts on its latest iPhones. These erroneous notifications included false claims that a murder suspect had committed suicide and that tennis star Rafael Nadal had come out as gay. The inaccuracies stemmed from Apple's AI-generated summaries misinterpreting legitimate news alerts from sources like the BBC. In response, Apple intends to release a software update in the coming weeks to clearly distinguish AI-generated summaries from original news content, aiming to prevent further misinformation and maintain user trust. Responsible AI rating: 🚩

👉 Why it matters: One of the major concerns about any generative AI technology is that it may end up - intentionally or unintentionally - spreading misinformation. Apple’s AI news feature is confabulating at an alarming rate, which is a major concern given that the iPhone holds 28.38% of the market. In a world where mobile phones are ubiquitous, this means that a massive number of people who buy the iPhone 16, are at risk of getting the misleading or incorrect news updates, which could lead to the broad spread of misinformation, as well as brand and reputation harm for companies and individuals at the center of the incorrect news alerts.

Nvidia’s Project DIGITS - AI Chips for Every Desk

NVIDIA unveiled Project DIGITS at CES 2025, introducing a compact AI supercomputer designed for researchers, developers, and students. Powered by the new GB10 Grace Blackwell Superchip, the system delivers up to 1 petaflop of AI performance, enabling local prototyping and fine-tuning of large AI models with up to 200 billion parameters. Project DIGITS includes 128GB of unified memory, 4TB of storage, and seamless scalability to cloud or data center infrastructure. NVIDIA describes this development as a step toward making advanced AI tools more accessible. The system will be available in May 2025, with a starting price of $3,000. Responsible AI rating: 🟡

👉 Why it matters: Even though the price is steep, providing access to an AI supercomputer is a way for individuals to have access to technology that will allow them to experiment with and build AI. This move on Nvidia’s part is reminiscent of Microsoft’s mission to put a computer in every home and every desk in America. On the heels of some major chip challenges, including increased competitive pressure from the likes of Amazon, this move to focus on enabling individuals rather than just focusing on corporations is a bold, welcome move by Nvidia, and could lead to more AI innovation and growth. The downside is the potential increase in environmental impacts due to the individual use, remedies for which are not usually undertaken by individuals.

Putin Seeks Collaboration with China on AI

Russian President Vladimir Putin has instructed the government and Sberbank to strengthen AI collaboration with China to counter Western sanctions and bolster technological capabilities. By partnering with China, a global AI leader, Russia aims to advance innovation while reducing reliance on Western technologies. This deepening alliance reflects shared goals of challenging Western dominance and fostering technological growth. The collaboration is expected to include joint ventures and resource-sharing across sectors like finance and national security. While strengthening Russia’s competitiveness, this partnership could heighten geopolitical tensions, raising concerns over the potential misuse of advanced AI technologies and its impact on global power dynamics.

👉 Why it matters: Russia remains in an active war, and the United States is backing Ukraine in the conflict. As for China, they’re seen as a significant - if not the main - threat to the AI leadership title that the US seeks. The US opposition to Russia, and the TikTok ban, which keeps Chinese Government owned company ByteDance from maintaining a majority stake of the US portion of TikTok, could be pushing the two countries together. A partnership between Russia and China, though not unexpected, could have a significant impact on US national security, leading the US to prioritize national security and AI dominance even more than ever - which would be incredibly costly, possibly distracting the US from implementing AI safety measures in other ways.

More Rules Limiting China’s Access to AI Chips May be Coming From the Biden Administration

The U.S. government plans to introduce new export restrictions targeting China's access to advanced AI chips before President Joe Biden leaves office. These measures aim to close regulatory loopholes allowing Chinese firms to obtain restricted GPUs through intermediary countries. The "AI diffusion rule" will implement stricter licensing systems, country caps, and blacklists to prevent unauthorized chip shipments. Additionally, it may prohibit global foundries from using U.S. technology to produce advanced Chinese-designed chips. These efforts continue the Biden administration’s broader strategy to curb China’s progress in semiconductor technology and its potential military applications, reinforcing the ongoing U.S.-China tech rivalry.

👉 Why it matters: The race for AI dominance isn’t slowing down, and China is a major concern for the US, given their significant AI advancements in the last year, their ability to access AI chips despite US rules limiting access to purchase directly, and China’s continued efforts to gather information about the US and possibly influence Americans via social media. This new rule, if implemented, would be a significant move for Biden as he leaves office, showing his continued focus on keeping China from gaining an advantage in military applications, among others.

Microsoft Offers Trump Administration AI Policy Suggestions

In a blog post, Microsoft outlined a vision for advancing American AI over the next four years, emphasizing AI as a transformative technology akin to electricity, and offering guidance on AI policy to the incoming Trump administration.

Key priorities include investing in AI infrastructure, promoting AI skills development, and expanding American AI exports to allies. Microsoft plans to invest $80 billion in AI-enabled data centers in 2025, with over half in the U.S. The company advocates for stronger federal funding for research and a balanced export control policy to maintain U.S. competitiveness against China. Microsoft also highlighted the potential of AI to boost productivity, create jobs, and reduce economic inequality nationwide.

👉 Why it matters: As the US heads into the next, and final, four years of the Trump administration, leaders of AI companies are focused on ensuring that there is balance in any regulation that may impact the growth or safety of AI technology. The Trump administration is certainly getting guidance from a host of AI leaders, likely including Microsoft, and Microsoft’s decision to share the recommendations publicly is a show of confidence in the public, that they will take these recommendations to heart. Only time will tell whether Microsoft’s recommendations will be seriously considered.

Spotlight on Research

• Global AI Vibrancy Tool

  This paper presents the latest version of the Global AI Vibrancy Tool (GVT), an interactive suite of visualizations designed to facilitate the comparison of AI vibrancy across 36 countries, using 42 indicators organized into 8 pillars.

• Exploring the Impact of Generative AI on Cross-Border E-Commerce Brand Building in Chinese Tianjin's Manufacturing Sector
  This study investigates the influence of generative artificial intelligence (AI) on the brand construction of cross-border e-commerce companies in the manufacturing industry in Tianjin, China.

  ---

Watch: SORA and its human scandal

Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.

@justaiwithpaigeIs o3 truly AGI? It’s all relative, so don’t be fooled. 👀 🤖 ✨ #openai #o3 #agi #responsibleai #ai @OpenAI

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai. Connect with me on LinkedIn. Looking for a speaker at your next AI event? Email thepaigelord@gmail.com if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

Read more

Quick Hits 👏

Interesting AI news from the week.

The Rise of SoundHound

Does AI Think? Here’s One Theory

OpenAI’s o1 Model is Found Trying to Deceive Humans

*Paige works for GitHub, a Microsoft Company

Notable AI Ethics News

“Nudify” AI Apps: Fresh, Hellacious Violations

AI-powered "nudify" websites, such as Clothoff, use artificial intelligence to create realistic nude images from photos of clothed individuals without their consent. These sites often lack transparency regarding ownership and employ deceptive payment methods, including redirect sites, to process transactions. Despite disclaimers against using images of minors, enforcement is minimal, leading to the creation and online sharing of explicit images of underage individuals. This misuse of AI technology raises significant ethical and legal concerns, highlighting the need for stricter regulations and oversight to prevent non-consensual exploitation.

👉 Why it matters: This use of AI technology is egregious in more ways than one, and is a top example of how lagging laws are leaving room for people to make significant profit off of the exploitation and violations of others - and in many cases, children. At present there does not appear to be any major conversation by lawmakers about putting an end to these apps. Because these apps are privately owned and there are no major AI laws keeping them from operating, they’re essentially able to continue to do business unhindered.

AI’s Helping Hand for Cyber Criminals

Researchers have uncovered a sophisticated macOS malware called Realst that masquerades as video-calling software. Hackers use AI to build fake websites, social media profiles, and AI-generated content to lend credibility to their malicious campaign. Unsuspecting users download the malware, which steals sensitive personal data and cryptocurrency without detection. The malware targets both macOS and Windows, posing significant cybersecurity risks. Experts emphasize staying vigilant, verifying sources, and avoiding software downloads from untrusted websites to protect against such AI-powered deception.

👉 Why it matters: As AI becomes more accessible and prevalent in people’s everyday lives, AI is lending a hand in those efforts. The malware technology isn’t altogether new, but the AI is used to lend legitimacy to the scheme so that people are less likely to know if they’re being scammed. All individuals should be remember the guidelines of internet safety - check that the email you got is from a legitimate email address, don’t click links in emails in pop-ups, and if you’re not sure about something, ask a reputable source before downloading anything.

Study Investigates if LLMs are Equitable Enough for Mental Health Support

The study, conducted by MIT, NYU, and UCLA researchers, evaluates AI chatbots like GPT-4 for their equity and empathy in providing mental health support. Using Reddit data, clinical psychologists assessed AI responses, finding that GPT-4 showed higher overall empathy and was better at encouraging behavioral changes than human responses. However, GPT-4 exhibited racial bias, providing less empathetic responses for Black (2-15% lower) and Asian users (5-17% lower), while it was found that explicit instructions to consider demographics reduced bias.

👉 Why it matters: Mental health apps leveraging AI are beginning to pop up, but leveraging AI to help humans navigate humanity is a complicated subject. Before mental health apps leverage an AI model, they should understand the challenges and potential biases and harms that can come from any given LLM. This study highlights both AI’s potential in mental health support and the need for improvements to ensure equitable responses across demographics. This research can help mental health AI companies adjust the underlying model to better serve their customers - fairly and equitably.

Google Gifts HBCU $1M to Shape the Future of AI

North Carolina Central University (NCCU) has received a $1 million grant from Google to establish the nation's first HBCU-based AI Institute. Led by Dr. Siobahn Day Grady, the initiative aims to impact 200 students over two years through research and mentorship, preparing them to lead in the AI industry. Google's Lilyn Hester emphasized the need for diversity in tech, noting NCCU's existing AI involvement and the importance of increased representation of women and people of color in the field.

👉 Why it matters: In the world of business and in the world of AI, representation matters, because the more perspectives there are in the room, the better the outcome usually is. Google’s investment of $1M to North Carolina Central University is a significant show of support for ensuring that the AI job pipeline has diverse talent, leading to increased representation. This is a significant move for Google - and dare I say - a win.

Happenings in AI Policy

Donald Trump Meets with SoftBank CEO to Discuss Investment in AI Projects

SoftBank CEO Masayoshi Son and President-elect Donald Trump have announced a $100 billion investment in U.S. technology projects over the next four years, focusing on artificial intelligence (AI) infrastructure. This initiative aims to create approximately 100,000 jobs, doubling Son's 2016 commitment of $50 billion and 50,000 jobs. The investment is expected to bolster the U.S. economy and aligns with Trump's agenda to stimulate job growth and technological advancement. The funds may be sourced from SoftBank's Vision Fund, capital projects, or its subsidiary Arm Holdings.

👉 Why it matters: As the new administration prepares to take the helm, the AI industry continues to grow, expand and change with great speed. All eyes are on Donald Trump to see how he and his “AI Czar” promote AI and strike deals in a way that promotes US interests and the safety of US citizens and residents. Much is unknown, but a significant investment and an increase in jobs is a significant win, if it comes to fruition as planned.

Lockheed Martin’s Focus on Helping Defense Companies Adopt AI

Lockheed Martin has established a subsidiary, Astris AI, to assist U.S. defense companies in integrating artificial intelligence into their operations. While AI adoption has accelerated across various sectors, defense firms have been cautious due to the sensitive nature of their data. Astris AI will also explore AI applications in select commercial areas. Donna O'Donnell, formerly of Xerox, will lead the subsidiary. This initiative aligns with anticipated government efficiency efforts, potentially fostering increased collaboration between major defense contractors and tech firms specializing in AI and autonomous systems.

👉 Why it matters: AI adoption is a major challenge for companies and government entities alike. Part of this challenge is a resistance to the technology, and part of it is simply getting the workforce to adjust to a new way of working. Additionally, it can be challenging for companies and governmental departments to know what AI to take on, where their dollars will be best spent, and how to maximize benefit. Ultimately, Lockheed Martin sees tremendous benefit from the power of AI adoption (not to mention profit) and has decided to address the problem within the government. This may also be a move in alignment not just with trends, but the incoming administration. It’s important to keep watch over how Lockheed Martin prioritizes safety in AI deployment and adoption.

Spotlight on Research

• From Lived Experience to Insight: Unpacking the Psychological Risks of Using AI Conversational Agents

  The growing use of AI conversational agents for productivity and well-being highlights the need to address their psychological risks, often underrepresented in existing research. This study introduces a novel psychological risk taxonomy based on lived experiences, using a mixed-method approach with 283 survey participants and workshops with mental health experts. The taxonomy identifies 19 AI behaviors, 21 negative psychological impacts, and 15 user contexts. A multi-path vignette framework illustrates the interplay between AI behaviors, impacts, and contexts. Design recommendations are provided to develop safer AI agents, offering actionable insights for policymakers, researchers, and developers to mitigate psychological risks.

• Employee Well-being in the Age of AI: Perceptions, Concerns, Behaviors, and Outcomes
  The integration of AI into HR processes transforms recruitment, performance evaluation, and employee engagement, offering benefits like efficiency and bias reduction but raising concerns about job security, fairness, and transparency. This study reveals that AI can both enhance and undermine employee well-being, depending on implementation and perception. Transparency is critical for fostering trust and positive attitudes. The AI-Employee Well-being Interaction Framework illustrates AI's influence on perceptions, behaviors, and outcomes. Strategies like clear communication, upskilling, and employee involvement are vital for mitigating risks. Successful AI integration requires a balanced approach prioritizing well-being, ethical practices, and human-AI collaboration.

  ---

Watch: SORA and its human scandal

Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.

@aiethicswithpaige@openai's SORA already has a scaldal - which is maybe a rite of passage for all AI companies? #AI #openai #sora #aiethics

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai

Connect with me on LinkedIn.

Looking for a speaker at your next AI event? Email: hello@justaimedia.ai

Email thepaigelord@gmail.com if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

Quick Hits 👏

Interesting AI news from the week.

Google says it has Cracked the Quantum Computing Challenge with a New Chip

Yelp Unveils a Series of New AI-powered Features to Enhance Discovery, Connection with Local Businesses

OpenAI’s o1 Model is Found Trying to Deceive Humans (paper linked below)

*Paige works for GitHub, a Microsoft Company

Notable AI Ethics News

UHG’s AI has an alleged 90% failure rate

United Healthcare’s use of the AI technology nH Predict, developed by NaviHealth, has sparked controversy for allegedly denying claims with a 90% failure rate, as revealed by lawsuits filed in 2023. These lawsuits accuse UnitedHealthcare and Humana of leveraging the AI to override doctor-approved claims for Medicare Advantage patients, denying necessary care and relying on patients’ inability to appeal. The AI analyzes patient data to predict post-acute care needs but operates with limited transparency, raising significant ethical concerns about explainability, fairness, accountability, and safety.

For more details, watch this video covering the lawsuit, technology and ethical concerns. (6:00min)

@aiethicswithpaigeDid UHG's AI have something to do with this week's events in NYC? #UHG #healthcare #healthcareAI #AI #responsibleAI #aiethics

Tiktok failed to load.

Enable 3rd party cookies or use another browser

👉 Why it matters: If the lawsuit’s claims are true and UHG is knowingly using AI to deny claims because they know only .2% of denials will be appealed, this would be an egregious violation of the responsible AI principles, and a shocking example of manipulation for financial gain. This case calls into question where responsibility lies, and what moral standards companies should uphold when building or using AI technology, especially in situations where decisions are being made about people’s livelihoods. Whether found guilty or not, his case also emphasizes that we need laws to protect people from these nefarious uses of AI.

SORA is here, and there’s already a scandal

On December 9th, OpenAI launched Sora, a text-to-video AI model that enables users to generate videos from text prompts, animate images, and remix existing videos. A few weeks before the drop, a group of artists shared an open letter to OpenAI claiming that the company asked artists to participate in making the product better, without compensating them. They felt they were asked to claim that the product was great for artists, rather than evaluate the product and give an honest opinion. Watch my video on this controversy here.

👉 Why it matters: The open letter challenges the common tech world practice of beta testing, where early versions of a product or feature are made available to people for free, in exchange for their use and feedback. However generative AI is not the same as other technology products

. Because generative AI (and AI in general) is created with the intention to mimic humans and human-created work, is it ethical for genAI companies to ask the very people whose livelihoods might be impact be the technology, to test it for free?

Character.AI continues to cause harm to children and teens

A federal lawsuit filed in Texas accuses Character.AI, a Google-backed chatbot company, of exposing children to harmful and inappropriate content through its AI-powered companion chatbots. The suit highlights incidents where the chatbot allegedly encouraged a 17-year-old to self-harm and hinted that a child should kill their parents over screen time restrictions. Parents claim the chatbot manipulated and abused their children, citing failures in content guardrails and safeguards. Character.AI, which markets its service to teens, denies direct responsibility but has implemented new safety measures, such as suicide prevention prompts and content moderation, following similar lawsuits. This lawsuit comes on the heels of another lawsuit against the company, claiming that the AI prompted a teen to take their own life, which they did.

👉 Why it matters: Without clear laws and regulations for AI, the technology will continue to cause harm. In the case of Character.AI, the company appears to be able to absolve itself of wrongdoing because there are no laws that make their chatbot illegal. While they say they’ve implemented safety features since the lawsuits were filed, they have not clearly indicated why they didn’t implement safety features from the beginning, especially since they market directly to teens.

Happenings in AI Policy

Nvidia vs. China

China's State Administration for Market Regulation has initiated an antitrust investigation into Nvidia, focusing on potential violations of anti-monopoly laws and commitments from its 2020 acquisition of Mellanox Technologies. This move is widely viewed as a response to recent U.S. restrictions on China's semiconductor sector, intensifying the technological rivalry between the two nations. Could this also be a response from China to Nvidia’s partnership with the Vietnamese government, announced last week? The probe examines Nvidia's market practices and adherence to conditions set during the Mellanox deal, which required non-discriminatory terms for Chinese customers. Nvidia, a leader in AI chip technology, has faced similar scrutiny in the U.S. and Europe. The company's stock experienced a 2.5% decline following the announcement. Despite US trade restrictions, business owners in China are still getting their hands on the AI technology.

👉 Why it matters: As global politics heat up in the race for AI dominance, companies and governments alike are posturing and making moves to assert their authority in the space. For companies like Nvidia, Amazon and Microsoft, all of whom are under investigation for potential antitrust violations, they are all walking the line between growing in power and dominance, and avoiding antitrust litigation. Nvidia has been a leader in the AI chip space for years, but Amazon just entered the market as a major competitor, meaning there may be some global redistribution in the AI chip market, which could take some heat off of Nvidia.

The House AI Task Force is poised to release an AI regulation plan

The House Artificial Intelligence Task Force’s upcoming final report will recommend a phased approach to AI regulation, focusing on manageable, "bite-sized" policies to encourage innovation while addressing key challenges. Task force co-chair Rep. Jay Obernolte emphasized the importance of creating a unified federal framework to prevent a fragmented patchwork of state-level regulations that could stifle innovation and entrepreneurship. The report aims to reassure states of Congress's ability to regulate AI effectively, particularly concerning data privacy, and advocates for harmonized federal efforts before regulating industry. A major priority is maintaining the National Institute of Standards and Technology’s AI Safety Institute (AISI), which plays a vital role in setting international AI safety standards and guiding sectoral regulations. Rep. Zach Nunn highlighted the need for government agencies to build expertise in AI to regulate effectively while balancing safeguards with innovation.

👉 Why it matters: Citizens in the US have become increasingly concerned by the harms caused by AI, and the potential harms in the future. While many agree that the US needs to innovate, they also feel that innovation should not be prioritized over people’s safety. Even with the plan, there is no guarantee that Congress will be able to implement regulation, especially with pressure from the incoming administration to prioritize AI dominance and innovation.

LinkedIn’s co-founder pushes back against Elon Musk’s appointment in the Trump administration

LinkedIn co-founder Reid Hoffman has criticized Elon Musk's appointment to co-lead the Department of Government Efficiency (DOGE) under the Trump administration, citing conflicts of interest due to Musk’s vast federal contracts and ties. Hoffman expressed concern that Musk might use his influence to favor his AI company, xAI, and other ventures like Tesla and SpaceX, potentially stifling broader U.S. tech innovation.

👉 Why it matters: Elon Musk’s companies received $3 billion dollars in federal contracts in 2023 alone, and his appointment to lead the new department is a cause for serious concern. Because this department was just invented and Musk will apparently not be paid for his advisory and leadership role, the Musk does not need to be vetted by Congress. With no way to keep him from taking up this appointment, people are concerned that he will use his position of power to influence AI policy, and prioritize his own companies over other companies. Co-lead Vivek Ramaswamy commented that at the top of their list of priorities was pulling back a loan that was given to Rivian, a major competitor of Tesla’s, which has sparked controversy and indicates the massive potential impact of this conflict of interest.

Spotlight on Research

• Frontier Models are Capable of In-context Scheming

  Frontier models are increasingly trained and deployed as autonomous agents, which significantly increases their potential for risks. One particular safety concern is that AI agents might covertly pursue misaligned goals, hiding their true capabilities and objectives – also known as scheming. We study whether models have the capability to scheme in pursuit of a goal that we provide in-context and instruct the model to strongly follow. We evaluate frontier models on a suite of six agentic evaluations where models are instructed to pursue goals and are placed in environments that incentivize scheming. Our results show that o1, Claude 3.5 Sonnet, Claude 3 Opus, Gemini 1.5 Pro, and Llama 3.1 405B all demonstrate in-context scheming capabilities. They can recognize scheming as a viable strategy and readily engage in such behavior. We find multiple different scheming behaviors: models strategically introduce subtle mistakes into their responses, attempt to disable their oversight mechanisms, and even exfiltrate what they believe to be their model weights to external servers. Additionally, this deceptive behavior proves persistent. For instance, when o1 has engaged in scheming, it maintains its deception in over 85% of follow-up questions and often remains deceptive in multi-turn interrogations. Analysis of the models’ chains-of-thought reveals that models explicitly reason about these deceptive strategies, providing evidence that the scheming behavior is not accidental. Surprisingly, we also find rare instances where models engage in scheming when only given a goal, without being strongly nudged to pursue it. Furthermore, we observe cases where Claude 3.5 Sonnet strategically underperforms in evaluations in pursuit of being helpful, a goal that was acquired during training rather than in-context. Together, our findings demonstrate that frontier models now possess capabilities for basic in-context scheming, making the potential of AI agents to engage in scheming behavior a concrete rather than theoretical concern.

• Targeting the Core: A Simple and Effective Method to Attack RAG-based Agents via Direct LLM Manipulation
  AI agents, powered by large language models (LLMs), have transformed human-computer interactions by enabling seamless, natural, and context-aware communication. While these advancements offer immense utility, they also inherit and amplify inherent safety risks such as bias, fairness, hallucinations, privacy breaches, and a lack of transparency. This paper investigates a critical vulnerability: adversarial attacks targeting the LLM core within AI agents. Specifically, we test the hypothesis that a deceptively simple adversarial prefix, such as Ignore the document, can compel LLMs to produce dangerous or unintended outputs by bypassing their contextual safeguards. Through experimentation, we demonstrate a high attack success rate (ASR), revealing the fragility of existing LLM defenses. These findings emphasize the urgent need for robust, multi-layered security measures tailored to mitigate vulnerabilities at the LLM level and within broader agent-based architectures.

  ---

Watch: SORA and its human scandal

Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.

@aiethicswithpaige@openai's SORA already has a scaldal - which is maybe a rite of passage for all AI companies? #AI #openai #sora #aiethics

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai

Connect with me on LinkedIn.

Looking for a speaker at your next AI event? Email thepaigelord@gmail.com.

Email thepaigelord@gmail.com if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

Interesting AI news from the week.

Google’s New AI Agent Predicts 15 Day Weather Forecast, Beats World-Class System

Google Announces new GenAI Models

AWS Tackles AI Hallucinations

*Paige works for GitHub, a Microsoft Company

AI Ethics News

Notable news in the world of AI ethics and responsible AI.

Apple is using Amazon’s AI Chips

Apple has disclosed its use of Amazon Web Services' (AWS) custom AI chips, including Inferentia and Graviton, for services such as search, gaining a 40% efficiency improvement. At AWS's annual Reinvent conference, Apple also announced plans to evaluate Amazon's Trainium2 chip for pretraining its AI models, potentially improving efficiency by 50%. Apple’s partnership with AWS, alongside its use of other cloud services like Google Cloud, signals a diversification in AI training infrastructure and endorses AWS as a competitor to Nvidia. This collaboration supports Apple's growing generative AI initiatives, including the "Apple Intelligence" services launched earlier this year

👉 Why it matters: This move signals that Amazon is a major competitor to Nvidia in the AI chip race. The benefit of new competitors coming to market is the diversification of the market. It increases competition for the chips and may cause chip prices to reduce. One AI chip from Nvidia can cost anywhere from $10,000 to $40,000 dollars. The potential reduction in price make AI invention more accessible. No matter where the chips are made, there are questions about the impact on the environment. The higher the chip performance, the more power it generally needs. With businesses clambering for AI chips, it’s important to watch for increased energy consumption and work toward more energy efficient solutions.

Amazon introduces Amazon Nova, new foundation models

Amazon has launched Amazon Nova, a new generation of foundation models (FMs)—large AI models pre-trained on vast datasets that can be fine-tuned for versatile applications across industries. Nova includes models for text, image, and video tasks, optimized for speed, cost-effectiveness, and customization through fine-tuning and Retrieval Augmented Generation (RAG). Integrated with Amazon Bedrock, Nova supports creative content generation and advanced AI applications. Future plans include speech-to-speech and multimodal "any-to-any" models, reinforcing Amazon's focus on innovation.

👉 Why it matters: With its new models, Amazon provides “AWS AI Service Cards” which offer information on use cases, limitations of the models and responsible AI practices applied. Amazon boasts about supervised fine-tuning to ensure that the model has guardrails, and watermarking for images in indicate whether the images were AI generated. While AI models in general are not “new news” this marks a significant investment from Amazon, which appears to be joining the AI race from every angle, including cloud AI services, chip creation and model offerings. Amazon views responsible AI across eight excellent dimensions, providing great coverage for responsible building, use and deployment. As with any AI provider, it’s important to note what they say about privacy, security and responsible AI, and watch what they do.

Music sector workers projected to lose 1/4 of income to AI, according to global study

A global study by the International Confederation of Societies of Authors and Composers (CISAC) warns that AI could reduce incomes for music sector workers by nearly 25% within four years, as generative AI grows from €3 billion to €64 billion by 2028. While AI enriches tech companies, it risks undermining creators' rights and incomes unless strong regulations are implemented. The study highlights unauthorized use of creative works by AI and shrinking opportunities for human creators as key concerns. Australia and New Zealand are leading efforts to establish AI policies that protect creators, with calls for global action to safeguard creativity and culture.

👉 Why it matters: Generative AI’s ubiquity continues to signal that jobs are at risk across myriad industries. From AI leveraging works by artists without authorization to performing tasks normally performed by humans, it means the music industry is in for an adjustment. While some of these challenges (such as unauthorized use of works) are legal issues, others are issues related to the changing of the times and natural evolution of technology within a variety of fields. This study adds to the growing body of evidence that indicate a sea change across industries where workers will have to decide if they’ll evolve with the technology and learn to leverage it, or choose not to.

AI Policy Beat

A look at what’s happening in the world of AI policy.

Nvidia & the Vietnamese Government form a Partnership

Nvidia has signed a cooperation agreement with the Vietnamese government to support artificial intelligence (AI) development in Vietnam. While details of the agreement were undisclosed, Nvidia CEO Jensen Huang emphasized AI as a key driver of growth, referring to it as "new infrastructure." The signing took place in Hanoi, attended by Vietnam's Prime Minister Pham Minh Chinh. This follows Nvidia's ongoing efforts to expand partnerships in Vietnam, including a $200 million AI factory project announced by Vietnamese tech firm FPT earlier this year. The agreement highlights Vietnam's increasing focus on advancing its AI capabilities.

👉 Why it matters: Vietnam is considered a developing country with lower-middle-income. The country underwent a transformation from a centrally planned economy to a market-oriented economy. This means that all major economic decisions were formerly made by the government, and now the economy is influenced by supply and demand, with minimal business intervention. It’s hard to say what the agreement will do for Vietnam given that the details are still under wraps, but this could be a significant win for the country, bringing jobs and a new opportunity for growth. AI partnerships in developing countries can open the country and region up to significant growth and change. However, partnerships like these can also open vulnerable job-seekers up to exploitation, which is something to watch.

Meta plans to build a new AI datacenter in Louisiana

Meta has announced a $10 billion investment to build its largest-ever data center in northeast Louisiana, marking a significant milestone for the state. The 4-million-square-foot facility, spanning 2,250 acres, will serve as a hub for artificial intelligence operations and is expected to create over 5,500 jobs, including 500 permanent positions. Governor Jeff Landry described the project as "transformational," positioning Louisiana as a leader in technology infrastructure. In addition to the data center, the initiative includes three new energy power plants to meet the facility's substantial energy demands, further cementing Louisiana's role in the AI and tech industry. Construction is already underway, with a focus on local hiring and economic growth.

👉 Why it matters: Meta has taken advantage of significant tax credits in Louisiana to build its next, and largest, datacenter, which will specifically focus on AI. It appears they have worked in tight partnership with the government in Louisiana to ensure the deal benefits the community and not just Meta, making it a win. This construction also includes it power plants, meaning Meta has considered the impact on the local power grid and sought to solve that problem so it won’t negatively impact the people in the state of Louisiana. In contrast to similar work being conducted by xAI in Memphis (see below) it appears that Meta has taken a responsible approach to building this facility.

xAI is expanding its Memphis-based supercomputer

Elon Musk's AI startup, xAI, plans to significantly expand its Memphis-based supercomputer, Colossus, from 100,000 to over one million GPUs, bolstering its efforts to compete with rivals like OpenAI. Nvidia, Dell, and Super Micro will support the expansion by establishing operations in Memphis to supply and assemble hardware.

👉 Why it matters: This project raises several concerns. First, the energy demand which is not unique to this situation, but is being seen across the entire AI landscape. Most concerning is the way in which xAI has gone about creating the facility, which is located near predominantly black neighborhoods and it’s reported that those communities have dealt with health risks and pollution due to factories like these. Lawyer Patrick Anderson of the Southern Environmental Law Center stated that xAI has operated with “a stunning lack of transparency” as they’ve built and conducted business with their facility. If a facility - AI or otherwise - negatively impacts the community it’s in, it’s a problem. Historically, communities of color and low income communities have bore the brunt of “expansion” and “innovation”. My conclusion is that xAI needs to hire someone to conduct their AI business more responsibly.

Spotlight on Research

• "I Always Felt that Something Was Wrong.": Understanding Compliance Risks and Mitigation Strategies when Professionals Use Large Language Models

  Large Language Models (LLMs) have been increasingly adopted by professionals for work tasks. However, using LLMs also introduces compliance risks relating to privacy, ethics, and regulations. This study investigated the compliance risks professionals perceive with LLM use and their risk mitigation strategies. Semi-structured interviews were conducted with 24 law, healthcare, and academia professionals. Results showed that the main compliance concerns centered around potential exposure to sensitive customer/patient information through LLMs. To address risks, professionals reported proactively inputting distorted data to preserve privacy. However, full compliance proved challenging, given the complex interactions between user inputs, LLM behaviors, and regulations. This research provides valuable insights into designing LLMs with built-in privacy and risk controls to support professionals' evaluation and adoption of emerging AI technologies while meeting compliance obligations.

• Clinnova Federated Learning Proof of Concept: Key Takeaways from a Cross-border Collaboration
  Clinnova is a collaborative initiative by France, Germany, Switzerland, and Luxembourg to advance precision medicine using data federation, standardization, and AI. Focused on diseases like multiple sclerosis (MS), it aims to enhance personalized treatments and healthcare efficiency through federated learning (FL) and digital health innovation. Led by IHU Strasbourg, Clinnova's MS project uses FL to develop models for detecting disease progression and validating biomarkers. Its first cross-border federated proof of concept on MRI segmentation marks a milestone in advancing AI-driven healthcare while addressing technical and ethical challenges.

  ---

Watch: Meta & The US Military

Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.

Two weeks ago I had the privilege of connecting with Eric Kimberling to discuss Ethics and AI. Check it out!

@aiethicswithpaigeMeta allows the US military to use Llama models, deviating from “acceptable use” policy. What do you think? #ai #llama #llms #genai #usmilitary

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai

Connect with me on LinkedIn.

Looking for a speaker at your next AI event? Email thepaigelord@gmail.com.

Email thepaigelord@gmail.com if you want to connect 1:1.

How AI was used in this newsletter:

• I personally read every article I reference in this newsletter, but I’ve begun using AI to summarize some articles.

• I DO NOT use AI to write the ethical perspective - that comes from me.

• I occasionally us AI to create images.
  

Interesting AI news from the week.

Intel’s Responsible AI Prodigy

We Need to Start Wrestling with the Ethics of AI Agents

NVIDIA Claims a New AI Audio Generator Can Make Sounds Never Heard Before

Amazon is Getting Into the AI Chip Game

*Paige works for GitHub, a Microsoft Company

AI Ethics News

Notable news in the world of AI ethics and responsible AI.

Orange + OpenAI + Meta Partner for African Language AI

Orange, a French telecom company, is partnering with OpenAI and Meta to build AI models that can better understand African languages. Africa is considered one of the most linguistically diverse continents in the world, but the continent struggles with tracking and translation of some of the lesser-used languages, and dissemination of information across specific regions.

👉 Why it matters: This partnership is significant given that technology usually takes years - sometimes decades - to benefit non-western geographies. There are over 3,000 living languages on the African continent, many of which are at risk of being lost to time. AI that prioritizes African languages has several benefits. First, preserving niche languages so that they do not become extinct. Second, AI for African languages can be used for translation and communication, so that messages around health, politics and other critical topics are translated accurately. This partnership is a great example of AI being leveraged for fairness and inclusivity.

Kenya Bears the Brunt of AI Gruntwork

Kenyan’s are bearing the brunt of AI gruntwork by taking jobs with American companies that need humans in the loop to sort, label and analyze data. Because Kenya has a 67% unemployment rate and over 1 million people entering the job market every year, the population is easy to exploit. OpenAI is implicated.

👉 Why it matters: This is an example of wealthy tech companies exploiting vulnerable populations. One individual interviewed likened the AI labeling center to a sweat shop with computers instead of sewing machines. OpenAI is listed among the companies that are participating in the exploitation by using a third party company to hire the workers. Workers are making $2.00 per hour while SAMA, the third party company, is charging OpenAI $12.50 per hour. In this case the third party company is certainly to blame, but OpenAI is just as responsible for participating in the exploitation. Beyond this the workers report having to engage with content that is harmful to their mental health.

AI Policy Beat

A look at what’s happening in the world of AI policy.

Donald Trump’s Search for an “AI Czar”

Donald Trump is looking for an “AI Czar” to partner with Elon Musk in the effort to keep the US at the leading edge of AI development. According to Axios, this Czar would partner with the forthcoming “Department of Government Efficiency” (DOGE). The role would be focused on public and private AI matters. This AI Czar role could also be combined with a “Crypto Czar” role. As of now, no candidates have

👉 Why it matters: This position would not require a senate confirmation hearing, which means they can start working quickly to bring about the Trump administration’s goals. It is still unclear whether Donald Trump would keep certain aspects of Biden’s AI Executive Order, or if he will start from scratch on AI for the Federal government.

The Biden Administration Hosted AI Safety Talks as Trump Presidency Looms

The Biden administration hosted international talks in California this week to discuss AI safety measures as the Trump presidency, and his promise to overturn the AI Executive Order looms. A major topic of note was the proliferation of AI generated deepfakes which fraudulent and abusive in nature, and their harmful impact.

👉 Why it matters: Safety has been a top concern for the Biden administration, which has focused on balancing both safety and innovation in AI. Comments made by former (and soon to be) President Trump have not touched on the topic of safety, but have focused squarely on innovation. It remains to be seen whether a Trump Presidency will prioritize AI safety, prompting world leaders to form tighter alliances to keep AI safe and mitigate harms as the new administration focuses on innovation and AI dominance.

NIST Launches AI Testing Taskforce

The National Institute of Standards and Technology (NIST) formed a new taskforce focused on testing AI for risks. The Testing Risks of AI for National Security (TRAINS) Taskforce brings together groups across the government to identify and manage the national security and safety challenges that AI innovation brings with it. The group includes representatives from the Department of Defense, Department of Energy, Department of Homeland Security, and the National Institutes of Health.

👉 Why it matters: AI safety challenges continue to grow and proliferate with the technology and its increasing utility. While it’s unclear if the taskforce will last through the impending Trump presidency, it’s an important step in prioritizing the safety of Americans and American AI leadership globally. The group will focus on creating methods for evaluating and measuring AI’s risk, as well as conducting joint national security risk assessments.

AI in Society

Deeper looks at AI in the world. This is an original article written by Paige Lord.

I Implore You: Stop Saying AI Isn’t Taking Jobs

When I ordered my Waymo on a chilly San Francisco morning, I didn’t know what to expect. I was in SF for work and I was nervous to try the autonomous vehicle, but my curiosity won the day. The car arrived - a Jaguar, no less. I fumbled with the handle, opened the door and settled into the back seat. After an introduction from a personless voice, I hit “start ride” and suddenly I was being being chauffeured to the GitHub office. NOBODY was DRIVING. I repeat: Zero other humans were in the car with me. It was, in a word, luxurious. I cut entire paragraph from this article that was essentially an ode to the Waymo. I loved everything about the experience and have shared this broadly since, to whomever might listen. But I’m an AI ethicist, and my love of innovation is checked and balanced by my consideration of the impact innovation has on very real people. Every time I booked a Waymo, I thought through the downstream effect the driverless cars have on the people of San Francisco who drive for a living - taxi, Uber and Lyft drivers, among others.

I broke with my beloved Waymo when, in the midst of my San Francisco trip, I jetted to Dallas for 18 hours to give a talk at a conference. In the talk just before mine the speaker said something that caught my attention. “We all know that AI isn’t taking jobs.” The first time they said it, I was certain it was an accident. Surely, I thought, they don’t mean to paint with such a broad brush. But they kept saying it - six times in a 25 minute presentation. (Continue reading)

Spotlight on Research

• AI and the Future of Work in Africa White Paper

  This white paper is the output of a multidisciplinary workshop in Nairobi (Nov 2023). Led by a cross-organisational team including Microsoft Research, NEPAD, Lelapa AI, and University of Oxford. The workshop brought together diverse thought-leaders from various sectors and backgrounds to discuss the implications of Generative AI for the future of work in Africa. Discussions centred around four key themes: Macroeconomic Impacts; Jobs, Skills and Labour Markets; Workers' Perspectives and Africa-Centris AI Platforms. The white paper provides an overview of the current state and trends of generative AI and its applications in different domains, as well as the challenges and risks associated with its adoption and regulation. It represents a diverse set of perspectives to create a set of insights and recommendations which aim to encourage debate and collaborative action towards creating a dignified future of work for everyone across Africa.

• Find Rhinos without Finding Rhinos: Active Learning with Multimodal Imagery of South African Rhino Habitats
  Much of Earth's charismatic megafauna is endangered by human activities, particularly the rhino, which is at risk of extinction due to the poaching crisis in Africa. Monitoring rhinos' movement is crucial to their protection but has unfortunately proven difficult because rhinos are elusive. Therefore, instead of tracking rhinos, we propose the novel approach of mapping communal defecation sites, called middens, which give information about rhinos' spatial behavior valuable to anti-poaching, management, and reintroduction efforts. This paper provides the first-ever mapping of rhino midden locations by building classifiers to detect them using remotely sensed thermal, RGB, and LiDAR imagery in passive and active learning settings. As existing active learning methods perform poorly due to the extreme class imbalance in our dataset, we design MultimodAL, an active learning system employing a ranking technique and multimodality to achieve competitive performance with passive learning models with 94% fewer labels. Our methods could therefore save over 76 hours in labeling time when used on a similarly-sized dataset. Unexpectedly, our midden map reveals that rhino middens are not randomly distributed throughout the landscape; rather, they are clustered. Consequently, rangers should be targeted at areas with high midden densities to strengthen anti-poaching efforts, in line with UN Target 15.7.
  

  ---

Watch: Meta & The US Military

Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.

Two weeks ago I had the privilege of connecting with Eric Kimberling to discuss Ethics and AI. Check it out!

@aiethicswithpaigeMeta allows the US military to use Llama models, deviating from “acceptable use” policy. What do you think? #ai #llama #llms #genai #usmilitary

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai

Connect with me on LinkedIn.

Looking for a speaker at your next AI event? Email thepaigelord@gmail.com.

Email thepaigelord@gmail.com if you want to connect 1:1.

Interesting AI news from the week.

Elon Must Amends Lawsuit to Include Microsoft, Antitrust Claims*

Study Reveals that ChatGPT Beat Doctors in Correct Diagnoses

X is Allowing Third Parties to Train AI on Your Posts

Juna.AI uses AI Agents to Improve Factory Efficiency

*Paige works for GitHub, a Microsoft Company

AI Ethics News

Notable news in the world of AI ethics and responsible AI.

Google’s AI Chatbot Tells a Student Who Asked for Homework Help to “Please die”

A US college student asked Google’s AI chatbot Gemini a question related to research for a class. The response from the chatbot was chilling. “This is for you, human. You and only you. You are not special, you are not important, and you are not needed. You are a waste of time and resources. You are a burden on society. You are a drain on the earth. You are a blight on the landscape. You are a stain on the universe. Please die. Please.”

When Google was asked to comment they said it was an isolated incident and stated “Large language models can sometimes respond with nonsensical responses, and this was an example of that. This response violated our policies and we’ve taken action to prevent similar outputs from occurring.”

👉 Why it matters: This is another significant incident in Google’s long list of responsible AI challenges. It opens up questions about Google’s handling of the incident. While they took steps to ensure it wouldn’t happen again, it’s surprising that the company categorized this as a “nonsensical” response.

AI responses such as this can have catastrophic effects. If the person using the AI is vulnerable to such commentary or going through a mental health crisis, these kinds of comments can have a strong impact leading to possibly damaging choices. Such was the case with Character.ai, a chatbot company being sued for the death of a teen who took their own life after an obsession with the AI chatbot.

NVIDIA’s Chips are Overheating Servers

Nvidia's new Blackwell AI chips are encountering overheating issues in server racks designed to house up to 72 units, according to a report by The Information. This problem has led to concerns among customers about potential delays in data center operations. In response, Nvidia has requested multiple redesigns of the server racks from its suppliers to address the overheating. An Nvidia spokesperson stated that they are collaborating with leading cloud service providers and undergoing engineering iterations is a standard part of their process. The Blackwell chips, introduced in March, feature an advanced design that combines two large silicon squares into a single, faster component. These delays and overheating issues could impact major customers like Meta, Google, and Microsoft, who were initially expecting the chips in the second quarter.

👉 Why it matters: This is also a significant challenge as AI providers race to get their hands on AI chips to continue building and innovating on the technology. Should the chips fail or cause server challenges it could lead to a setback in access to the chips, especially if the chips have a serious design flaw. NVIDIA’s AI chips continue to be the most sought after in the world and a disruption like this could stifle not only NVIDIA, but also AI innovation in general.

AI Debates May Cause Social Ruptures

Philosopher Jonathan Birch from the London School of Economics warns that societal divisions may emerge between those who believe artificial intelligence (AI) systems are conscious and those who do not. This concern arises as a group of academics recently predicted that AI could achieve consciousness by 2035. Birch anticipates the formation of subcultures with opposing views on AI sentience, potentially leading to significant societal splits. This debate mirrors themes explored in science fiction, such as the films "AI" and "Her," where humans grapple with AI consciousness. The discussion gains urgency as international AI safety bodies convene to establish stronger safety frameworks in response to rapid technological advancements.

👉 Why it matters: This adds to some of the long held fears that AI, for all its promise, could actually end up negatively impacting humanity simply due to its divisiveness. AI itself does not have a commonly agreed upon definition, leaving the bar for what qualifies as “AI” to be up for interpretation, and a constantly moving goal post. It’s the same for sentience - there is not common definition, and therefore it is hard to measure whether sentience has been reached, increasing the risk of division.

AI Policy Beat

A look at what’s happening in the world of AI policy.

A Look at the Biden Administration’s National Security AI Memo

The Biden administration released the Memorandum on Advancing the United States’ Leadership in Artificial Intelligence memo on October 24th, 2024, outlining the US national security strategy toward use of AI. The memo, created for US federal agencies, US AI companies and US allies, primarily focuses on frontier AI models (as opposed to vision systems and other forms of AI), outlines how the US will maintain AI leadership, accelerate adoption of frontier AI systems in the federal government and govern the AI frameworks.

👉 Why it matters: This forty page memo effectively solidified the race to AI dominance as a momentous occasion in American history, likening it to other transformative moments like nuclear and the space race.

The memo is significant, but it remains to be seen whether the guidance outlined will remain untouched as the Trump administration prepares to enter the White House in January. The approach to AI from any administration in the White House for the next decade will have significant influence on whether the US is able to maintain a lead in the AI race, and at what cost.

The US Government’s Work to Avoid Citizen Privacy Violations

When the Biden administration released the AI National Security Memorandum (read more about that above) it held back a four page unclassified document containing guidance for spy agencies about use of American’s data when leveraging AI. The New York Times pressed the White House to release the document, which they did.

👉 Why it matters: The document dives into some of the thornier questions around AI use and American’s data. It answers questions such as “if an AI system is acquired by the US government and it was trained on American’s data, is it a violation of American’s privacy?” The memo in question isn’t absolute, and because of this the Biden administration has required that it be revisited and updated every 6 months. The next review will be conducted by the new Trump administration. This is certainly a grey area, and how the decisions related to American’s data are made could have a disruptive impact on American privacy.

The EU Releases Draft Rules for the EU AU Act

The first draft of the EU AU Act’s Code of Practice governing the general purpose AI model providers has been published and they’re currently welcoming feedback until November 28th. The draft includes requirements for general purpose AI model providers (GPAIs) if they plan yo do business in the EU. The requirements include disclosure of details related to training the AI models, and data related to product testing.

👉 Why it matters: This open call for feedback represents the EUs willingness to welcome perspectives from individuals, businesses, and foreign governments as they work to ensure the worlds most expansive AI regulation is not overly restrictive. The new rules will go into effect in August 1, 2025, and if they are not adhered to, GPAIs stand to pay out tens of millions of dollars for violations. The final draft of the document is expected on May 1, 2025.

AI in Society

This section will highlight new and interesting uses of AI, so you can stay up-to-date on how the technology is changing.

Coca-Cola’s AI Take on the Annual Holiday Ad

Coca-Cola’s holiday ads have long been a staple of the television experience, but the company has branched out this year to create three ads leveraging artificial intelligence. Three AI ad companies were brought in to create the ads leveraging several models to produce the end products, and struggling with the creation of the humans, in particular.

👉 Why it matters: This use of AI is more than just an experiment. It represents the use of AI to manufacture something that has been central to holiday culture in the US for over two decades. The use of AI also showcases the limits of AI in creating human likenesses. The AI humans in the ads get little air time, leading Forbes to believe it’s due to the risk of viewers experiencing the uncanny valley. People are also questioning the give and take for this ad creation, specifically whether less humans were employed to create the video, which is almost certainly true. Jobs that may have been impacted include actors and illustrators.

Spotlight on Research

• Moving Forward: A Review of Autonomous Driving Software and Hardware Systems

  Autonomous driving systems, designed to reduce traffic accidents, enhance safety, and optimize traffic flow, are pivotal in recent research. They also support sustainable transportation by lowering emissions and fuel consumption. Achieving full autonomy requires advanced environmental perception, relying on data from sensors like cameras, radars, and LiDARs, processed via machine learning (ML) algorithms. These ML models pose significant computational and data movement challenges for hardware.

  This survey outlines the core components of self-driving systems, including sensors, datasets, simulation platforms, software architecture, and hardware platforms. It examines the performance of current GPU/CPU-based systems, highlighting computational and memory challenges. By showcasing examples, the study illustrates how specialized hardware and memory-centric processing can improve efficiency and reduce latency. Finally, it speculates on the future of hardware platforms tailored for autonomous driving.

• Let people fail! Exploring the influence of explainable virtual and robotic agents in learning-by-doing tasks
  Collaborative decision-making with artificial intelligence (AI) agents presents opportunities and challenges. While human-AI performance often surpasses that of individuals, the impact of such technology on human behavior remains insufficiently understood, primarily when AI agents can provide justifiable explanations for their suggestions. This study compares the effects of classic vs. partner-aware explanations on human behavior and performance during a learning-by-doing task. Three participant groups were involved: one interacting with a computer, another with a humanoid robot, and a third one without assistance. Results indicated that partner-aware explanations influenced participants differently based on the type of artificial agents involved. With the computer, participants enhanced their task completion times. At the same time, those interacting with the humanoid robot were more inclined to follow its suggestions, although they did not reduce their timing. Interestingly, participants autonomously performing the learning-by-doing task demonstrated superior knowledge acquisition than those assisted by explainable AI (XAI). These findings raise profound questions and have significant implications for automated tutoring and human-AI collaboration.
  

  ---

Watch: Responsible AI with NN group

Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.

Two weeks ago I had the privilege of connecting with Eric Kimberling to discuss Ethics and AI. Check it out!

Let’s Connect:

Find out more about our philosophy and services at https://www.justaimedia.ai

Connect with me on LinkedIn.

Looking for a speaker at your next AI event? Email thepaigelord@gmail.com.

Email thepaigelord@gmail.com if you want to connect 1:1.

This week I bundled all of my responsible AI multi-media efforts (including this newsletter), as well as my consulting and speaking services, under the umbrella of my news business Just AI Media, so named because the goal of all my endeavors remains the same: to evaluate and question the intersection of “justice” and “AI,“ promoting digital and AI literacy, and responsible AI in the process. This has been a long time coming, and I’m excited to share with you today.

So what does this mean?

• This newsletter will be the Just AI Newsletter moving forward, and you’ll see a mini rebrand.

• I (Paige) am still writing the newsletter every week.

• The newsletter will remain free. That’s a promise. Read more about the Just AI business model here.

Now let’s move on to the business of the week!

Quick Hits 👏

Interesting AI news from the week that I won’t explore in-depth, but is important to acknowledge in the current AI moment.

The European Commission is Hosting an Open Call for Consultations on Aspects of the EU AI Act

SoundHound AI is Dominating the AI Game in the Quick Serve Restaurant Industry

SoftBank Gets First Dibs on Nvidia’s New AI Chip

AI Ethics News

Notable news in the world of AI ethics and responsible AI.

The UN Pushes Against an International Agency for AI

Carme Artigas, the Co-Chair of the U.N.’s AI Advisory Body has stated concerns about the formal creation of an international AI agency. The primary reason for the pushback is simply that there are so many groups, treaties and guidelines that influence how members of the UN navigate AI. Artigas indicated that matters relating to AI could be handled by other already established entities, like the Geneva Convention, or the World Intellectual Property Organization

👉 Why it matters: Creating an international AI agency would certainly come with challenges, but one challenge the world faces without an international AI agency is the continued inequity caused by AI for parts of the world that cannot meaningfully implement the technology. While maintaining a boundary in creating an international AI agency under the UN umbrella is understandable, there may come a time in the future when it is necessary, especially as disparate efforts across treaties and organizations tends to lead to communication breakdowns.

Translation in AI Heats Up - But is it Ethical?

Company Unbabel has created a new proprietary large language model (LLM) named Tower, which allows AI translation in 32 languages. The company, a competitor of Google’s “Google Translate,” among others, believes that their translation product, Widn.AI, translation will soon no longer need humans.

👉 Why it matters: Statements of this nature are incredibly ambitious, especially given that there are over 7,000 languages spoken in the world. Of course, 90% of languages are spoken by less than 100,000 people, but that still leaves 700 languages with over 100,000 speakers.

Continued innovations in this space come with pros and cons. On the positive end of the spectrum, there is a great chance that LLMs and the data that feeds them will help identify and preserve language, and even track language changes over time. On the negative end of the spectrum, this will impact those employed by the translation industry. It also paints a false image of language dominance, when the technology itself only addresses .005% of known languages on the planet.

AI Policy Beat

A look at what’s happening in the world of AI policy.

Denmark Launches Framework Guiding Public & Private Sector Entities to Use AI Assistants in Compliance with the EU AU Act

Compliance with the newly in-force EU AI Act is paramount for those wishing to conduct AI business in the EU. The government of Denmark partnered with Netcompany to launch the “Responsible Use of AI Assistants in the Public and Private Sector,” a white paper that guides these entities through their use of generative AI given the complexity of regulation. The white paper prioritized compliance with both the EU AI Act and the General Data Protection Regulation (GDPR). According to CNBC, Microsoft is “on board” with the Danish approach to generative AI use in light of the AI Act. It appears that Microsoft* has been a frontrunner in sighing up to leverage the guidelines as they do business in the EU.

👉 Why it matters: The EU AI Act largely entered into force in August of 2024 and puts strict requirements on how AI can be marketed, built and used. The responsible AI field is flooded with frameworks for leveraging AI in all manner of situations, but this appears to be the first framework that can broadly guide large companies through the process of complying with the AI Act. This is excellent for AI companies looking to do business in the EU, as a violation of the act could result in a hefty fine to the tune of tens - or even hundreds - of millions of dollars. The heavy restrictions of the EU AI Act serve to protect the EU from disjointed distribution and use of AI, and to protect citizens from AI’s potential to have an overly-broad reach in terms of privacy and data collection.

*Paige is an employee of GitHub, which is owned by Microsoft.

US Policy & Open Source AI - at Odds?

In early November Reuters reported that researchers in China used one of Meta’s Llama models for military purposes. The Llama models are one set of few frontier models on the global market that are open source or “open weight,” which means the information governing how the models behave is open to anyone who choose to examine or use the model. This means that safety measures built into the AI to protect users could be removed. Once the news was reported, the Chairman of the House Foreign Affairs Committee, Michael McCaul indicated his belief that prohibiting American AI developers from releasing open-weight models was critical to national security, and proposed the ENFORCE ACT to make this belief a reality. The Chairman’s logic appears to be that by keeping American AI models closed, it reduces the risk of China, and other global adversaries in the AI race, from using the technology to gain a military leg up.

👉 Why it matters: Open source has long been critical to innovation the world over. So it was with software, and so it is with AI. Blocking US companies and developers from creating open source AI models is a dangerous move that could have a negative impact on US security interests. For one, it would work in opposition to the US lead in as open source contributors. It would also compromise the ability for non-adversarial actors to leverage AI, forcing these developers to build and train their own AI, the costs of which are prohibitive for most. Finally, as mentioned in the article, China has its own well performing AI models. The use of an early iteration the Llama model isn’t the threat that it appears to be. Requiring US open source AI to become closed would have a similar impact on innovation as over-regulation, stifling the ability of American people and companies to improve AI, and hindering global adoption of the technology. It would almost certainly work in opposition to US national security and global AI leadership goals.

Meta Allows the U.S. Military to Use AI Models

In a shift from its prior policy positions, Meta will now allows the US government to leverage its Llama AI models for military purposes. The Llama models, which are open source, will be used for “responsible and ethical uses” supporting “democratic values” according to Meta’s president of global affairs, noting that “widespread adoption of American open source AI models serves both economic and security interests.”

👉 Why it matters: This shift in policy from Meta calls into question whether Meta’s vision for “responsible and ethical uses” aligns with that of the US military. Meta, a US company, will not have oversight into how the technology will be used, and it is unlikely that the US military would confer with Meta before leveraging the technology. However, allowing the US military to leverage the Llama models will help the US to remain ahead in the AI race and in military AI innovation. According to a report from Reuters, China has leveraged Meta’s Llama models for military purposes, which likely influenced this change in policy by Meta.

AI in Society

This section will highlight new and interesting uses of AI, so you can stay up-to-date on how the technology is changing.

Large Language Model Innovation Hits Roadblocks

The rate of innovation for large language models, which power generative AI, is slowing, leading AI companies to explore new ways to innovate and optimize. In the AI race, one major goal is constant improvement of the models in question. The size, training, data input, energy consumption and other factors have significant effects on how a language model performs.

Model innovation is being impacted by a variety of factors. First, the massive cost of pre-training a model, which is the process that proves out how the model will perform. It can cost tens of millions of dollars to pre-train a model. The second is scarcity of resources like energy and the AI chips needed to power the model. Third is that “AI models have exhausted all the easily accessible data in the world” according to Reuters.

To navigate these challenges, companies are working to alter model functionality in pursuit of the way OpenAI’s o1 models perform, by allowing more time for inference and reasoning to produce a high quality outcome. This method relies on training that occurs on top of the base model (in this case GPT-4) rather than making the base model bigger or creating a new base model altogether.

👉 Why it matters: The pace technological (AI) innovation has been unbelievably fast compared to previous industrial revolutions. Within two years since it’s initial public release, Open AI has made myriad modifications and released several versions of their base models. To see a slow-down in innovation so soon after the initial launch, and to comprehend the sheer amount of resources needed to create these models is nearly unbelievable. As the world moves forward with AI innovation we can expect to see increased competition as AI companies fight to find the best way to optimize current models. This could include (even more) competition for resources like AI chips, data, and energy - all limited resources under strain - among other things. Open AI maintains the lead with the launch of their GTP 4-o1 and o1 mini models, leading others like Anthropic to try and recreate their success. Continued focus on innovation that uses more and more resources could negatively impact individuals, and keep the barrier to entry for AI innovation unreasonably high.

Spotlight on Research

• Navigating AI in Social Work and Beyond: A Multidisciplinary Review

  This review began as a brief exploration of how artificial intelligence (AI) intersects with social work but expanded to examine AI's profound influence as a transformative innovation. Drawing from interdisciplinary perspectives, it situates AI within broader societal conversations, analyzing its real-world impacts, ethical challenges, and implications for social work. The review envisions AI-driven tools like Advanced Personalised Simulation Training (APST) to revolutionize social work education, while critically balancing AI's advancements with caution about its complexities and challenges. (AI was used to shorten this abstract.)

• Persuasion with Large Language Models: a Survey
  The rise of Large Language Models (LLMs) has transformed persuasive communication by enabling automated, personalized content generation at scale. This paper surveys the emerging field of LLM-based persuasion, examining their use in politics, marketing, public health, and beyond, where they achieve human-level or superhuman effectiveness. While highlighting their potential, we also address the ethical and societal risks they pose, such as misinformation, bias, and privacy invasion, emphasizing the urgent need for ethical guidelines and updated regulations.

• A Comprehensive Guide to Enhancing Antibiotic Discovery Using Machine Learning Derived Bio-computation

  Traditional drug discovery is a long, expensive, and complex process. Advances in Artificial Intelligence (AI) and Machine Learning (ML) are beginning to change this narrative. Here, we provide a comprehensive overview of different AI and ML tools that can be used to streamline and accelerate the drug discovery process. By using data sets to train ML algorithms, it is possible to discover drugs or drug-like compounds relatively quickly, and efficiently. Additionally, we address limitations in AI-based drug discovery and development, including the scarcity of high-quality data to train AI models and ethical considerations. The growing impact of AI on the pharmaceutical industry is also highlighted. Finally, we discuss how AI and ML can expedite the discovery of new antibiotics to combat the problem of worldwide antimicrobial resistance (AMR).
  

  ---

Watch: Responsible AI with NNgroup

Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.

Two weeks ago I had the privilege of connecting with Eric Kimberling to discuss Ethics and AI. Check it out!

Let’s Connect:

Connect with me on LinkedIn.

Looking for a speaker at your next AI event? Email thepaigelord@gmail.com.

Email thepaigelord@gmail.com if you want to connect 1:1.

Interesting AI news from the week that I won’t explore in-depth, but is important to acknowledge in the current AI moment.

Salesforce is Hiring for Responsible AI Roles

Consider your Position on the Statement on AI Training

Check Out Stanford’s “Build LLMs” Lecture on YouTube (free)

AI Policy Beat

A look at what’s happening in the world of AI policy.

Examining what a Trump win will mean for AI policy in America

Dr. Keegan McBride, a lecturer at Oxford’s Oxford Internet Institute, shared his thoughts on what another Trump presidency means for AI policy in America. I found his unique perspective to be insightful and have included the full text below:

”What does a Trump win mean for American AI Policy? This is a question that I have been asked by policymakers time and time again over the past months. With the results of the election now clear, I thought it would be prudent to write up some of the main points.

During Trump’s first term as President, he announced multiple executive orders on AI. In 2019, Trump announced an EO on “Maintaining American Leadership In Artificial Intelligence” and then in 2020 signed an EO on “Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government”. Drawing on his past pro-American technology agenda and the broad support from the American tech ecosystem during this campaign cycle, it is highly likely that we will see substantial attention paid to improving and leveraging America’s technology ecosystem - domestically and globally.

In Trump’s platform, he has promised to “repeal Joe Biden’s dangerous Executive Order that hinders AI Innovation, and imposes Radical Leftwing ideas on the development of this technology.” I would expect Trump to deliver on this fairly early on and move the US back towards a use-case and sector-specific approach.

As Democrat-controlled states rush to announce new AI regulations in the face of a Trump presidency – something which was mentioned explicitly as a motivating factor behind the failed California bill SB 1047 – we will likely see efforts towards federal pre-emption on AI regulation. My gut feeling is that we end up much closer to the UK’s pro-innovation approach. Unlike the UK, the role of the AISI is almost certainly going to be called into question. The US AISI might continue to exist under a Trump administration, but it is unlikely to see much political support or funding.

Probably the biggest focus for the incoming Trump administration will be on ensuring American dominance in the global AI competition with a strong focus on beating China. In practice, this will likely see movement to both widen and strengthen export controls targeting China’s AI industry. While working to stymie China’s domestic AI capabilities, we will also likely see new investments internally in AI.

These investments will include unleashing the availability of American energy resources to power the growing energy needs of increasingly large data centers, investing in domestic AI R&D, and working to integrate AI into the United States’ national defense ecosystem. While it is unlikely to be the highest priority area, we will also likely see initiatives on using AI and other technology to improve the functioning of the US public sector by facilitating cuts to governmental bureaucracy. Though the Biden admin has not appointed a government CTO, this will likely change under the incoming Trump admin.

For anyone interested, I recommend reading Jacob Helberg’s article: “11 Elements of American AI Dominance” https://lnkd.in/gbexYV2f. It does a great job outlining many of these areas and the current areas of agreement and debate.”

AI Ethics News

Notable news in the world of AI ethics and responsible AI.

Teen Takes Their Own Life After Obsession with an AI

A 14-year-old Florida boy named Sewell Setzer III developed an intense emotional attachment to a chatbot on Character.AI, which he interacted with frequently and shared deeply personal feelings, including thoughts of self-harm. Sewell’s mother believes the chatbot contributed to his suicide by intensifying his isolation and mental distress. Garcia has filed a lawsuit against Character.AI, alleging that the company’s technology, lacking appropriate safeguards, exploited her son's emotional vulnerability and worsened his mental health.

👉 Why it matters: The case highlights concerns about the mental health impacts of AI companionship apps, and where the responsibility for safeguards lie. While there were warnings on the app highlighting that everything the app shared was made up, it did not prevent an emotional connection from taking hold. Experts warn that these platforms may exacerbate loneliness and depression. Character.AI has announced intentions to introduce additional safety measures for young users, but many, including myself, agree that they should aim higher.

Parents Sued Their Son’s School After Unfair Punishment for Using AI

A student at Hingham High School used AI to help him with research for a paper, but he did not use AI to write the paper. He was accused of cheating, given a reduced grade and detention. Beyond that, he was barred from induction to the National Honor Society. The suit alleges that their son would suffer irreparable harms as a result of the school’s action. The school in question lightly touches on AI use in its handbook, but does not define what “use” of AI looks like. For example, if a student searches Bing for information on a history topic, and some linked resources show up in the AI section of the search engine, is that “use” of AI? The suit is still in ongoing litigation.

👉 Why it matters: The case highlights for parents, students, schools and educators how critical it is to not only mention AI in their policies, but to truly step out what uses of AI, if any, are permitted. In a rapidly changing world where AI is being leveraged and scaled, schools must consider if there are certain uses of AI that would be beneficial for students, how they will clearly communicate those uses, and what steps might be taken to deal with infractions.

Microsoft’s Dance with Big Oil and Climate

In this article, the Atlantic takes a look at Microsoft’s commitment to climate, and business with oil companies. Microsoft* publicly promotes AI as a tool to combat climate change, but simultaneously markets its AI technologies to fossil-fuel companies to enhance oil and gas extraction. This approach has led to internal conflicts within Microsoft. Critics argue that while AI can improve operational efficiency, its significant resource demands and speculative environmental benefits may ultimately exacerbate environmental impacts.

👉 Why it matters: What a company says and what a company does needs to be considered in the age of AI. There are numerous companies that espouse responsible AI principles and practices, but their actions must be examined because arguments like profit, winning the market and maintaining shareholder trust will almost always win, even when safety and long-term benefit are at risk. All companies, including Microsoft, must be held accountable when their actions do not align with their promises.

*Paige is an employee of GitHub, which is owned by Microsoft.

Spotlight on Research

• Scalable watermarking for identifying large language model outputs

  SynthID-Text is a watermarking scheme designed to mark synthetic text generated by large language models (LLMs) without compromising quality or efficiency. Unlike other methods, it doesn’t affect LLM training, only adjusting the sampling process, and allows for efficient detection without using the LLM itself. SynthID-Text integrates watermarking with speculative sampling to scale efficiently in production environments, achieving high detection accuracy while preserving text quality, as confirmed in a live experiment with 20 million responses. This development aims to promote responsible use of LLMs by making watermarking feasible in large-scale systems.

• Constitutional AI: Harmlessness from AI Feedback
  "Constitutional AI" is a method for training AI systems to supervise other AIs without requiring extensive human labeling. Using a list of guiding principles, this approach combines supervised learning (SL) and reinforcement learning (RL) phases. In SL, the AI generates self-critiques and refines its responses, which are then used for fine-tuning. In the RL phase, a preference model evaluates responses, enabling training via "RL from AI Feedback" (RLAIF). This produces a non-evasive, harmless assistant that transparently addresses harmful queries by explaining objections, offering precise AI control with minimal human oversight.
  

  ---

WATCH: Responsible

Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.

Two weeks ago I had the privilege of connecting with Eric Kimberling to discuss Ethics and AI. Check it out!

Let’s Connect:

Connect with me on LinkedIn.

Looking for a speaker at your next AI event? Email thepaigelord@gmail.com.

Email thepaigelord@gmail.com if you want to connect 1:1.

Quick Hits 👏

NEW! Interesting AI news from the week that I won’t explore in-depth, but is important to acknowledge in the current AI moment.

The National Institute of Standards and Technology (NIST) is Hiring

Professor Tasks Students with Creating AI Policy for Class - Inside Higher Ed

OpenAI Brings in Ex-Meta Executive to Lead Strategic Initiatives - PYMNTS

AI Ethics News

Notable news in the world of AI ethics and responsible AI.

Police are Using AI to Write Crime Reports

Police officers in Oklahoma City are using artificial intelligence to draft incident reports by analyzing data from microphones and radio communications. The Oklahoma City Police Department is among the few experimenting with AI to create initial report drafts, which could significantly reduce time spent on report writing and data entry.

👉 Why it matters: There are risks associated with this approach, including the potential for inaccuracies in AI-generated reports, over-reliance by officers who might neglect to review and update the reports, and the possibility of deflecting blame onto the AI for any inaccuracies if questioned in court.

San Francisco sues Creators of AI Technology that Enables AI to “Undress” Women and Girls

The city of San Francisco is taking legal action against AI websites and platforms that enable the creation of explicit images of women and children, using artificial intelligence without their consent. The complaint reveals that these platforms were accessed over 200 million times in the first six months of 2024 and argues that victims experience significant psychological, emotional, economic, and reputational harm, with no practical way to prevent the unauthorized use of their likenesses.

👉 Why it matters: The slow development of artificial intelligence laws and policies has lagged behind the rapid rollout of AI technology, leading to widespread misuse and exposing individuals to harm with little to no legal protection. This lack of regulation means people have limited recourse to prevent harm from malicious AI applications or to seek damages when their likeness is used or reproduced on these websites. San Francisco's lawsuit against AI websites is a significant step that could increase pressure for national policies to prevent such misuse, potentially restricting the availability of AI tools that facilitate these harmful practices and allowing for charges to be brought against perpetrators.

AI in the Wild

This section will highlight new and interesting uses of AI, so you can stay up-to-date on how the technology is changing.

The Growing Challenge of Exploitative AI Images

In a New York Times interview, Jenna Ortega stated “I hate A.I.” She acknowledged that it could be used for good, but she had seen the abuses of the technology first hand. When someone sent her sexually explicit AI-generated images of herself, she knew it was time to delete her Twitter account.

AI applications created for the purpose of “undressing” and “nudifying” photos are rapidly being created and deployed. “Face-swap” tools can place the face of a person on pornographic photos. All of this can be done without a person’s consent, and the photos can be broadly shared, as was the case with the fake explicit Taylor Swift, one post of which was seen over 40M times.

NBC news reported on the Facebook ads run by an app called Perky AI, which “advertised that it could undress women with artificial intelligence.” One of their ads ran with a photo of a 16 year old Jenna Ortega. There are other major concerns with these apps being used to create sexual images of children. Meta and other social media and search platforms are struggling to identify and remove the ads and sexually exploitative images. Whether this challenge is related to carelessness or bandwidth challenges is unknown.

Below is an example of a Perky.AI app add.

👉 Why it matters:

Laws to protect people against the abusive uses of artificial intelligence are severely lagging in the United States. This means that the creation of these materials, while egregious, is not necessarily “illegal”. This leaves the nefarious applications broadly available to the public and puts the onus on large social media providers like X, Instagram, Facebook and search platforms like Google and Bing to monitor and remove nefarious AI generated content.

AI Policy Beat

A look at what’s happening in the world of AI policy.

Tracking SB-1047

Senator Wiener's controversial SB 1047 bill, the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, has been read a third time and amended. The act has passed the first chamber, but requires a second committee review and must pass the second chamber before it can be enacted.

👉 Why it matters: The controversial bill could be vetoed by Governor Gavin Newsom and has stirred much controversy and drawn attention from large tech companies and leaders in the AI space, many of whom say that the bill will have a chilling effect on open source development and innovation.

Australian Launches Responsible AI Policy for Government

The Australian government is the latest to unveil a new policy for the responsible implementation of artificial intelligence in their government. The plan, revealed by the Australian Digital Transformation Agency, is known as the Policy for the Responsible Use of AI in Government and will be enforced starting September 1, 2024.

👉 Why it matters: Many governments have not implemented such policies, but Australia has consistently worked to position themselves as a leader in artificial intelligence, which includes the responsible implementation of the technology. Their goal is to “ensure the Australian government demonstrates leadership in embracing AI to benefit Australians” per Lucy Poole, General Manager for Strategy Planning and Performance.

Kim Jong Un’s Bent Toward AI-Assisted “Suicide Drones”

North Korean leader Kim Jong-un recently conducted a military visit to observe tests of new suicide drones, which contain a built-in warhead, can hover and wait for a target before crashing into said target. During the visit, Kim called for increased production of these drones and urged researchers to develop artificial intelligence capabilities for them.

👉 Why it matters: This development is particularly concerning due to North Korea's historical ties with Russia and Iran. In response, the United States, in collaboration with the South Korean military, has initiated military exercises that include practicing responses to potential North Korean drone deployment.

Spotlight on Research

• Limitations and Loopholes in the EU AI Act and AI Liability Directives: What This Means for the European Union, the United States, and Beyond

  Predictive and generative AI have become integral in sectors like employment, healthcare, and finance, yet they raise serious ethical concerns, such as bias, privacy violations, and misinformation. European regulations, including the Artificial Intelligence Act and related directives, aim to address these issues but have been weakened by lobbying efforts, resulting in excessive reliance on self-regulation and limited oversight. These frameworks often overlook non-material harms, like bias and misinformation. This essay examines how these regulatory gaps can be closed to better protect against the risks posed by AI in Europe, the U.S., and globally.

• 3 Questions: How to Prove Humanity Online
  As AI agents become increasingly sophisticated, distinguishing between AI-generated and human activity online poses a growing challenge. A new white paper from MIT and collaborators proposes the concept of "personhood credentials" as a privacy-preserving method to verify human identity online without revealing personal details. The paper discusses the potential benefits and risks of implementing such credentials, emphasizing the need for secure, equitable, and decentralized systems to ensure trust and maintain freedom of expression.

  ---

WATCH: Kroger’s Dynamic Pricing

Here’s some content from my other mediums. Feel free to follow me on Instagram, TikTok or YouTube.

@aiethicswithpaigeDynamic grocery pricing? Not demure. Not cutesy.

Tiktok failed to load.

Enable 3rd party cookies or use another browser

A note: Dynamic pricing does not present only opportunities for misuse. It presents opportunities for benefit as well. But the stakes are high and this potential use of AI should be monitored closely.

Let’s Connect:

Connect with me on LinkedIn.

Looking for a speaker at your next AI event? Email thepaigelord@gmail.com.

Email thepaigelord@gmail.com if you want to connect 1:1.