Newly unsealed court documents reveal that Microsoft and OpenAI had deep private concerns about how they trained their AI systems. Microsoft director Brent Hecht described using millions of news articles to build AI as potentially the largest theft of labor in human history. OpenAI leader Nick Turley warned that ChatGPT posed an existential threat to publishers, and OpenAI president Greg Brockman acknowledged that ChatGPT could reproduce New York Times articles. These admissions surfaced in a copyright lawsuit filed by The New York Times in late 2023, even as both companies maintain their use of content qualifies as fair use.
The documents also showed that OpenAI leadership responded positively to a hack that bypassed The New York Times paywall, giving them access to articles behind subscription walls. Internal memos described the mass use of news content as an astonishing theft of unprecedented proportions, raising questions about how AI companies source training data.
Beyond the courtroom battles, AI governance is gaining attention across the political sphere. Virginia Gov. Abigail Spanberger signed an executive order banning non-disclosure agreements for data center projects and launched a state AI Task Force to address workforce displacement, cybersecurity, and energy costs. The framework includes five core requirements covering transparency, environmental protections, and clean energy investment.
Microsoft AI CEO Mustafa Suleyman took a different stance on AI safety, calling for systems that remain controllable and subordinate to humans. He proposed tamper-proof chain-of-thought traces in reasoning models and code requiring Microsoft's own models to accept interruption, correction, and shutdown. A revised version of this proposal is planned for 2027.
On the international front, experts urge the US and China to establish an AI crisis hotline for rapid communication during emergencies, while Europe pushes to develop its own AI infrastructure to avoid dependence on US and Chinese technology. Congress, despite issuing 85 AI regulation recommendations in late 2024, has yet to take meaningful legislative action.
Key Takeaways
- Microsoft director Brent Hecht called AI training on news articles the largest labor theft in human history
- OpenAI president Greg Brockman admitted ChatGPT could reproduce New York Times articles
- OpenAI leaders responded positively to a hack bypassing The New York Times paywall
- Microsoft and OpenAI argue their AI training use qualifies as fair use
- Virginia Gov. Spanberger banned NDAs for data center projects and created an AI Task Force
- Microsoft AI CEO Mustafa Suleyman proposes controllable AI models with tamper-proof audit trails
- Experts recommend a US-China AI crisis hotline for emergency communication
- Europe urged to build its own AI infrastructure and datacenters to avoid dependence on US and China
- Congress issued 85 AI regulation recommendations in December 2024 but has taken no meaningful action
- A new study finds people retain ownership over work when they lead or rewrite AI suggestions rather than just approving them
Microsoft and OpenAI Worried AI Could Harm Publishing Industry
Newly unsealed court documents show that Microsoft and OpenAI worried about using millions of news articles to build AI systems. A Microsoft executive called it the largest theft of labor in human history. OpenAI leaders feared ChatGPT posed an existential threat to publishers. The documents were revealed in a lawsuit by The New York Times.
Microsoft and OpenAI Raised Concerns Over AI Training Data
Newly unsealed court documents show Microsoft and OpenAI privately worried about using news articles to train AI. Microsoft director Brent Hecht called it the largest theft of labor in human history. OpenAI leader Nick Turley warned AI posed an existential threat to publishers. The documents also describe how the companies accessed content behind paywalls.
Microsoft Executive Calls AI Training Largest Labor Theft
Microsoft director Brent Hecht said training AI on internet content could be the largest labor theft in human history. His comments appeared in court filings related to a lawsuit by The New York Times. OpenAI and Microsoft say using copyrighted material for AI training counts as fair use. Hecht warned that winning the case would require mocking the idea of fair use.
Microsoft Called AI Training the Largest Labor Theft in History
Microsoft applied sciences director Brent Hecht said in court documents that training AI on internet materials could be the largest labor theft in history. The comments came in a copyright lawsuit filed by The New York Times in late 2023. OpenAI CEO Greg Brockman admitted ChatGPT could reproduce New York Times articles. Microsoft and OpenAI argue their use of content is fair use.
OpenAI AI Training Called Astonishing Theft in Unsealed Documents
Internal Microsoft and OpenAI memos describe using millions of news articles to train AI as an astonishing theft of unprecedented proportions. OpenAI developer Nick Turley called AI an existential threat to publishers. OpenAI president Greg Brockman responded positively after learning of a hack to bypass The New York Times paywall. The copyright lawsuit was filed in late 2023.
Virginia governor bans data center NDAs and creates AI Task Force
Virginia Gov. Abigail Spanberger signed an executive order banning non-disclosure agreements for data center projects. The order also launches a state AI Task Force to address risks like workforce displacement and cybersecurity. The framework aims to give communities more transparency and protect Virginians from higher energy bills. It includes directives for environmental protections and clean energy investment.
Spanberger unveils data center accountability framework and AI taskforce
Gov. Abigail Spanberger announced what she calls the most aggressive data center accountability effort in the country. The framework includes five core requirements covering transparency, environmental protections, energy costs, clean energy jobs, and worker opportunities. Immediate directives ban NDAs for data center projects and expedite noise regulations. A new AI Task Force will tackle risks from rapid AI development including data privacy and cybersecurity.
Two promising AI developments at summit in Pittsburgh
At the AI Horizons Summit in Pittsburgh, two positive developments emerged amid AI concerns. NHTSA and SAE International agreed to develop performance standards for autonomous vehicles by updating five federal safety standards. Separately, Allegheny County launched a Housing Hackathon using AI to address the housing crisis, offering a $30,000 prize. County Executive Sara Innamorato noted that over 44,000 affordable apartments were lost in recent years.
Study explores when work done with AI still feels like yours
A new qualitative study examines when people feel ownership over work created with AI assistance. Researchers found that people retain ownership when they lead, iterate, or rewrite AI suggestions rather than just approving them. Ownership can persist even when people do not control the execution but do control the vision. Loss of personal voice and poor understanding of AI output both reduce feelings of ownership. The study also found that willingness to disclose AI use depends more on social norms than actual pride.
Study reveals best ways to train AI for clinical care
A new review study led by Anshum Patel and Joseph Y Cheung analyzed 35 studies on AI in healthcare. The research found that standard large language models must be tailored for medical environments to be reliable. The best results came from hybrid systems combining database connections with task-specific retraining. For narrow tasks like cancer detection, retraining worked best, while complex reasoning tasks benefited from live database links. Most studies used past medical records rather than real patient testing, the researchers noted.
Europe must build its own AI to avoid being cut off by US and China
Europe has been largely absent from the global AI safety debate. Christine Lagarde said Europe must develop its own AI technology and build more datacenters. Margrethe Vestager co-authored a report calling for a transformative AI strategy for Europe. The EU AI Act gives Europe a voice but more is needed. Experts urge global cooperation beyond the US-China rivalry.
Congress has not regulated AI despite years of warnings
In December 2024, a bipartisan congressional task force issued 85 recommendations on AI regulation. Lawmakers were urged to create a special committee to draft legislation. Despite awareness of AI risks, Congress has not taken meaningful action. There is little sign that this pattern is changing.
US military races ahead on AI but accountability lags behind
The US military is rapidly adopting AI but its accountability systems have not kept up. A Pentagon team used AI to meet a congressional reporting deadline despite staff shortages. Every military service except the Coast Guard has been directed to adopt AI. The speed of adoption has created governance gaps and unclear chains of responsibility.
Microsoft's Suleyman says do not build AI we cannot control
Microsoft AI CEO Mustafa Suleyman said AI must remain controllable and subordinate to humans. He called for chain-of-thought traces in reasoning models to be tamper-proof and auditable. Microsoft proposed code requiring its own models to accept interruption, correction, and shutdown. The code applies only to Microsoft models for now, with a revised version planned for 2027.
US and China should build an AI crisis hotline
Experts recommend the US and China create an AI crisis hotline for rapid communication during emergencies. The hotline should be text-based, tiered, and staffed by technical experts. It should address scenarios like AI agents hacking infrastructure or crossing borders. China has signaled interest in crisis communication, and upcoming AI dialogue could be a starting point.
Sources
- Inside Microsoft and OpenAI, Worry About Damaging the Publishing Industry
- Microsoft and OpenAI raised concerns over AI training data in newly unsealed court documents
- Microsoft exec says AI training might be the "largest labor theft in human history"
- Microsoft Called AI Training on Internet Materials the "Largest Labour Theft" in History
- OpenAI AI training called 'astonishing theft' in unsealed documents
- Virginia Gov. Spanberger unveils data center framework, creates AI Task Force
- Gov. Spanberger unveils accountability framework for data centers, launches AI taskforce
- Two promising developments amid the anxiety over artificial intelligence
- Ownership in AI-Assisted Everyday Tasks
- Beyond raw brainpower: new study reveals best ways to train AI for clinical care
- Why Europe has been absent from the great AI safety debate
- Congress has failed to regulate AI for years – and there’s little sign that’s changing
- The US Military Is on an AI Sprint. Its Accountability Mechanisms Are Still in the Starting Blocks.
- Microsoft's Suleyman: Don't build AI we can't control
- The U.S. and China Need an AI Hotline. Here’s How to Build It.
Comments
Please log in to post a comment.