A senior Microsoft executive admitted that AIs are “hoovering up” the work of millions of people, in a process that could be seen as “an astonishing theft of unprecedented proportions”.

The admission was contained in a cache of documents released by a US court on Thursday, from the ongoing legal spat between US newspapers and AI companies including OpenAI and Microsoft.

and the group are chasing the ChatGPT creator and its corporate partner over what they claim is the wholesale copying of millions of newspaper articles to fuel the training and operation of AI chatbots.

The unsealed – but partly redacted – legal documents set out the newspapers’ arguments and extensively quote the tech companies’ internal conversations. Most of the material is presented in a summary of the plaintiffs’ case – the underlying exhibits remain sealed, so they lack their full context.

They reveal the tech companies were worried they were creating a “doom loop” in which AI would make newspapers redundant, destroying the very information they copy to make their products useful.

Here are some of the revelations in the unsealed records.

The plaintiffs quote Microsoft’s own Director of Applied Science saying: “Millions of people around the world will soon consider large [AI] models’ hoovering up all their work to be an astonishing theft of unprecedented proportions… Almost no one intended for content they created to be used in this fashion, nor are they compensated for its use.”

The documents allege the person called it perhaps the “largest theft of labor in human history”.

The defendants argued that training AI on copyright news articles was “fair use”, which is allowable under law – but the newspapers quoted a Microsoft executive admitting that allowing copying on this scale would “make a complete mockery of the idea of fair use”.

OpenAI researchers discussed in 2023 that ChatGPT was able to “access the content” behind a paywall, including for “new york times too,” describing this as “great to avoid expensive subscriptions” but “not great for any company that provides a subscription-based service”.

One employee told OpenAI’s Co-Founder and President, Greg Brockman about “a hack to get around nytimes paywall” to help with Brockman’s efforts to “scrape” The Times’s site. Brockman allegedly responded: “Ah nice”.

OpenAI Policy Director Jack Clark is quoted saying: “[O]ur work on AI and Creativity is going to increasingly lead to us creating systems that substitute for the labor of the people that define the ‘culture’ of society.”

OpenAI’s Head of ChatGPT wrote that publishers face an “existential threat” from AI because people will increasingly use AI to find information that they used to read from the original source.

Brockman is quoted as saying that AI models are “particularly good at predicting text of news articles,” “excellent at news,” and “very good at any news task”.

In 2022, OpenAI employees acknowledged that ChatGPT would have “memorised a ton of data and therefore will be insanely good at regurgitation”.

They were so worried that it would “accidentally” regurgitate an article it had copied that they created a “filter” to stop AIs from outputting articles copied from the companies that were suing OpenAI. Microsoft’s Director of Applied Science, Dr Brent Hecht, allegedly called this an “accidental cover up”.

One internal Microsoft document, said in its legal brief, recognised that the race to fuel AIs with ever more content, then feed that content to its users, could drive news sites out of business – and thereby starve the AIs of fresh content.

“Our AI content strategy has started a ‘doom loop’ that will hurt the performance of our [AI] models and the entire web at the same time,” the Microsoft document says.

“It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM [large language model AI] business with respect to its ‘content supply chain’.”

One Microsoft document is quoted saying: “LLMs are a product that destroys its supply chain.”

The legal documents reveal that “doom loop” is already in action: users of its Copilot “answer engine” were up to 93 per cent less likely to click through to a newspaper website compared to a traditional search engine.

Microsoft’s CEO Satya Nadella agreed under oath that conversing with chatbots “has substituted … giving you the information right there on the website on the AI platform versus needing to go to the underlying source”.

OpenAI internal data as of February 2026 tracked that “the demand for reliable local news is already visible inside ChatGPT at a rate of about 1 million prompts per week”. The newspapers say the real number is far higher than that.

Even when an AI-generated response acknowledges the original source with a link, one OpenAI software engineer said: “No matter how prominently we show the links, users won’t click.”

argues that AI enables “the generation of high volumes of ‘pink slime’, which is unreliable low-quality content that is often either plagiarised or remixed from legitimate news sources”.

“[AI] products have made pink slime news content farms cheaper, easier to make, and harder to detect,” they say: it would cost less than $7,000 (£5,230) to generate a million 500-word news articles using plagiarised content.

This undermines the business model of news publishers, diluting the market with a “flood of low quality inauthentic news material”. gives the example of Prism News, which operates 200 AI-generated publications posing as local newsrooms and hobby sites, with only four employees.

The newspapers accuse the AI giants of breaching copyright on their content to “exploit the works for their intrinsic value and offer commercial products that supplant those works”.

They say AI companies remain powerless to break out of their “doom loop” because, “while the industry as a whole would benefit if every company paid to sustain the continued production of the creative works their technology depends on, each individual company is better off taking content for free while others pay.”

They quote OpenAI co-founder Greg Brockman writing he was “deeply motivated by the gazillions” he hoped to gain by commercialising OpenAI’s technology.

In a defence memorandum, also released on Thursday, OpenAI says newspapers have “no right” to stop their language models from learning new facts and telling users about them.

“Nobody owns facts, just as nobody owns language, grammar, or style,” they argue. “These technologies put the world’s facts within reach of anyone who asks, in whatever form is most useful to them. That is progress, and copyright law does not stand in its way.”

They also say that since ChatGPT was released, traffic to the website has continued to grow, its revenues have “increased substantially” and it has hit record-high subscriber numbers. Though advertising revenue is falling for many publishers, the tech companies argue that the trend started before AI came along.