Senior staff at the biggest AI companies privately warned that their products posed an "existential threat" to news publishers, even as their employers argued in court that the products do no such thing.
The revelation, flagged by The Information, the San Francisco subscription publication that covers technology and business, comes from court filings unsealed this week in a copyright case brought by The New York Times and other news organisations.
What the filings say
The lawsuit accuses OpenAI, the maker of ChatGPT, and Microsoft, its largest financial backer, of training their models on millions of articles without permission.
Newly unredacted passages quote Nick Turley, OpenAI's head of ChatGPT, telling colleagues that publishers faced an "existential threat" because AI products were becoming a substitute for the journalism they draw on.
A Microsoft director, Brent Hecht, went further, describing the mass scraping of news content as the "largest theft of labor in human history".
An internal Microsoft document warned that the company's own approach risked a "doom loop" that would damage both its models and the wider web.
Why it matters
The wording matters because it strikes at the companies' legal defence.
OpenAI and Microsoft argue that training AI on copyrighted articles is fair use, on the grounds that it transforms the material and does not compete with the original journalism.
The publishers say the internal warnings show the opposite, that the companies knew their tools would replace the very sources they relied on.
They also cite Microsoft's own traffic figures, which showed click-through rates from its Bing chatbot were between 87% and 93% lower for the Times's websites than from ordinary Bing searches.
Microsoft's chief executive, Satya Nadella, testified that chatbots had substituted for visits to publishers' sites, and that paywalled material should be licensed rather than simply taken.
The wider fight
The filings land in the middle of a year-long struggle between news organisations and AI firms.
Publishers say AI answer engines have drained the referral traffic and advertising revenue that funded their journalism, and they have responded on two fronts.
Some, like the Times, are pressing lawsuits, while others have signed multiyear licensing deals that turn lost readers into direct payments.
The Department of Justice has filed a statement supporting the tech companies' position, a sign of how contested the underlying legal question remains.
What happens next
The immediate result is a two-track publisher strategy of courtroom pressure and paid deals.
How the fair-use question is settled will decide whether AI systems can keep ingesting journalism freely or must pay for it, and that in turn will shape how news reaches readers in the months ahead.
