Newly unsealed court documents have shed fresh light on internal concerns at Microsoft and OpenAI over the use of millions of news articles to train artificial intelligence systems, including warnings that the technology could undermine the news organizations whose journalism helped develop the models.
The disclosures emerged Thursday in the copyright battle brought by The New York Times and other publishers against OpenAI and Microsoft in federal court in New York. The publishers are seeking summary judgment, arguing that evidence uncovered during litigation establishes that the companies copied and used copyrighted journalism without authorization.
Among the most striking revelations was an internal assessment attributed to Microsoft Director of Applied Science Brent Hecht, who described the large-scale use of online content to train AI models as “an astonishing theft of unprecedented proportions” and potentially the “largest theft of labor in human history.”
Microsoft has emphasized that Hecht's remarks represented his personal views rather than the company's position.
“These comments reflect one employee’s individual perspective, are not a legal analysis, and do not represent the company’s views,” a Microsoft spokesperson said.
The documents cited by the publishers also describe internal discussions about the effect AI-generated answers could have on traffic to news websites and the broader economics of journalism.
Millions of articles at the center of dispute
The Times and its co-plaintiffs allege that OpenAI used millions of news articles in developing its AI systems, including material from The New York Times and numerous other publishers.
The litigation, initially filed in 2023, has grown into one of the most closely watched copyright disputes involving generative AI. Publishers including Ziff Davis, The Intercept, the New York Daily News and other news organizations have pursued claims against OpenAI and Microsoft over the alleged unauthorized use of their work.
The publishers contend that the companies' AI systems were trained on journalism without permission and can subsequently provide information that competes with the original publications.
OpenAI and Microsoft dispute those claims and argue that their use of copyrighted material is protected by the legal doctrine of fair use. OpenAI has also maintained that the Times' claims have been narrowed during the litigation and has publicly defended its position in court filings.
The question of whether training AI models on copyrighted material constitutes fair use remains a major unresolved legal issue in the United States.
OpenAI employees warned of threat to publishers
The newly unsealed material goes beyond the amount of content allegedly used for training and focuses on concerns within the companies about the consequences for the media industry.
According to the court filing, Nick Turley, who led the team developing ChatGPT, warned in a 2023 internal memo that AI represented an “existential threat” to publishers. In another communication, he reportedly wrote that AI products would become increasingly “substitutive” as the technology improved.
The filings also cite Microsoft documents warning that the companies' AI strategy could create a “doom loop” in which declining traffic to news websites would weaken the broader online ecosystem and ultimately affect the quality of information available to AI systems.
The concern was particularly significant because news organizations depend heavily on visits to their websites, subscriptions and advertising revenue to finance journalism.
Microsoft CEO Satya Nadella also addressed the issue during a deposition, according to the filing. He acknowledged that AI systems could provide information directly to users rather than requiring them to visit the original publisher's website.
Debate over links and attribution
News organizations have long argued that generative AI systems could divert readers away from publishers by answering questions directly rather than sending users to the websites where the underlying reporting was produced.
AI companies have attempted to address the concern by displaying citations and links alongside answers. But the newly unsealed documents indicate that even some people inside OpenAI questioned whether links would be enough to preserve traffic to publishers.
“No matter how prominently we show the links, users won't click,” one OpenAI engineer said, according to the court document.
The publishers argue that such evidence is relevant to the fair-use question because U.S. copyright law considers, among other factors, whether a challenged use harms the market for the original work.
Dispute over paywalled material
The filings also contain allegations concerning the collection of material from news websites, including content protected by paywalls.
According to the publishers' account of evidence obtained during discovery, OpenAI employees discussed methods for obtaining New York Times content behind a paywall. The filing cites an exchange in which an OpenAI researcher mentioned a “hack to get around nytimes paywall,” to which OpenAI co-founder and president Greg Brockman responded, “ah nice.”
The companies' handling of paywalled material has become particularly significant in the case.
Nadella testified that paywalled content should be licensed for AI training or grounding and said that, had he known OpenAI had scraped and trained on such material, he would have exercised Microsoft's contractual rights concerning retraining, according to the unsealed filing.
Legal battle could shape AI's future
The case has implications well beyond the relationship between the technology and news industries.
OpenAI and Microsoft argue that AI training is a transformative use of copyrighted material and therefore falls within fair-use protections. The publishers, by contrast, argue that the systems were built using their work without permission and that AI products can compete directly with the original journalism.
The U.S. Department of Justice recently filed a brief supporting OpenAI and Microsoft, invoking considerations including scientific progress, economic growth and national security, according to the information provided in the case. The companies' broader legal position remains that their use of copyrighted material is protected under fair-use principles.
The Times and its fellow plaintiffs have asked U.S. District Judge Sidney Stein to resolve the case through summary judgment. If the judge grants the request, the relevant claims would be decided without a full trial; a final ruling is not expected until 2027.
The court has not ruled that OpenAI or Microsoft infringed the publishers' copyrights. The newly unsealed documents instead reveal evidence and internal communications that the publishers say support their claims, while the technology companies continue to contest the allegations and defend their use of copyrighted material under existing law.