Community Trust ScoreVerified
Unsealed court documents dropped a bombshell. Microsoft employees, in internal memos, questioned whether OpenAI’s use of news articles amounted to “the largest theft of labor in human history.” The documents came out as Judge Sidney Stein of the Southern District of New York weighs summary judgment motions tied to a lawsuit the New York Times filed in 2023 — a case that has since grown to include eleven other publishers.
The core allegation is pretty straightforward: Microsoft and OpenAI used paywalled content to train AI models without authorization. What makes the newly public documents so damaging isn’t just the legal claim — it’s what the companies’ own employees were saying behind closed doors. The internal memos paint a picture of people inside both organizations who understood, at least privately, that something was off about the way their AI systems were being built.
Nadella Testimony and the Paywall Problem
Microsoft CEO Satya Nadella testified that any use of paywalled content should be licensed. He went further, saying that if he had known about OpenAI’s practices, he would have pushed to retrain the AI models entirely. That’s a significant admission from the top, even if Microsoft has tried to soften the blow by clarifying that internal memos critical of AI practices were written by Brent Hecht — someone hired specifically to offer diverse perspectives, not someone speaking for the company officially.
But Hecht’s memos still exist. And they still said what they said. One of them flagged concerns about AI models potentially “destroying their supply chains” — meaning the very publishers whose content was being scraped to build these systems. Hard to walk that back.
OpenAI’s own staff weren’t exactly quiet about it either. Internal conversations showed employees discussing techniques to get around the New York Times paywall. Greg Brockman, the company’s president, acknowledged in one of those conversations that what they were doing amounted to a “hack.” That word is now sitting in the court record.
ChatGPT Team Saw Publishers as Collateral Damage
Nick Turley, who led the ChatGPT team at OpenAI, was blunt about it internally. He described AI products as an “existential threat” to publishers and flagged that these products were becoming increasingly substitutive — meaning they were starting to replace the original sources rather than complement them. A 2023 memo from Turley made clear he saw the trajectory and it wasn’t good for traditional media.
And the traffic argument? Basically gutted. An internal OpenAI memo acknowledged that AI chatbots probably won’t drive users back to news sites even when source links are displayed prominently. An OpenAI engineer went further, noting that despite those links being visible, users were unlikely to click through. That’s a direct hit to one of the main arguments AI companies use to justify their relationship with publishers — that they send traffic, that they’re partners, not predators.
Jack Clark, who was OpenAI’s policy director before leaving to co-found Anthropic, wrote a memo in 2020 warning company leaders about the broader societal fallout from AI. He saw AI systems as potential replacements for significant chunks of human cultural labor. He also worried, in fairly stark terms, about AI becoming a symbol of Silicon Valley barging into every corner of life and leaving a mess behind. Anthropic declined to comment on the documents.
The Fair Use Defense and What’s at Stake
Both Microsoft and OpenAI are leaning hard on fair use. Their argument is that training AI on news articles transforms the original works into something new and distinct — and that transformation is what makes it legal. Steven Lieberman, who represents several of the plaintiff publishers, pushed back on that framing, saying the unsealed documents show both companies’ own internal doubts about whether what they were doing was actually fair.
The New York Times declined to comment. OpenAI didn’t respond to requests for comment either. So the public record right now is mostly the internal communications — and those communications are not flattering.
The case has real stakes beyond just these two companies. Publishers across the industry have been wrestling with how to handle AI systems scraping their content, and the legal framework around fair use for AI training is still genuinely murky. A ruling from Judge Stein either way will probably shape how AI developers approach content licensing for years.
What’s clear from the documents is that people inside OpenAI and Microsoft weren’t operating in complete ignorance. They had conversations. They wrote memos. They used words like “hack” and “existential threat” and “largest theft of labor in human history.” Those words are now public, and they’re sitting in front of a federal judge.
Turley’s 2023 memo sits at the center of the publisher argument — AI products get more substitutive as they improve, which means the damage to traditional media compounds over time rather than stabilizing.
Frequently Asked Questions
Who filed the lawsuit against Microsoft and OpenAI over AI training data?
The New York Times filed the original lawsuit in 2023, and eleven other publishers later joined the case, which is being heard by Judge Sidney Stein in the Southern District of New York.
What did Satya Nadella say about using paywalled content for AI training?
Nadella testified that paywalled content should be licensed, and said he would have pushed for retraining AI models had he known about OpenAI’s practices.
Why It Matters
This controversy highlights the growing tension between technology companies and content creators regarding the ethical use of data in training AI models. As legal battles unfold, they could set significant precedents that may reshape the AI landscape, potentially affecting how companies like Microsoft and OpenAI access and utilize information, which in turn could influence market strategies and innovation in the AI sector. The outcome of this case may also impact investor confidence in AI technologies, depending on how intellectual property rights are interpreted in the context of AI development.
