Newly unsealed court filings in the New York Times‘ copyright lawsuit against OpenAI and Microsoft show the two companies’ own executives describing what an internal Microsoft document called an AI doom loop: a feedback effect in which generative products trained on scraped web content are, in the company’s own words, killing “the entire web” that supplied their training data.
The filing was unsealed on Thursday. It was written by lawyers for the New York Times as part of a motion for summary judgment, a request asking the court to rule without a full trial. Much of it consists of quotes pulled from Microsoft and OpenAI documents and depositions that had previously been sealed or redacted at the two companies’ own request. Read together, the quotes amount to an admission of what critics of large language models have argued for years: that the models were trained on copyrighted material without consent or payment, and that the people building them understood the risk this posed to the web that supplied it.
Microsoft’s own words describe an AI doom loop
The filing quotes an internal Microsoft document warning that millions of people would come to see LLMs “hoovering up” their work as an act of theft, and that the company knew it.
- The document said users would see this as “an astonishing theft of unprecedented proportions,” adding that “almost no one intended for content they created to be used in this fashion, nor are they compensated for its use.”
- The same material describes the scraping behind LLM training as “the largest theft of labor in human history.”
- Other language cited in the filing says the practice makes “a complete mockery of the idea of fair use,” the legal doctrine OpenAI and Microsoft have relied on to justify training on copyrighted text without a licence.
OpenAI, which has already ruled out an IPO for 2026, is defending this case alongside Microsoft.
Microsoft distances itself from its own scientist
Several of the most pointed lines in the filing come from Brent Hecht, Microsoft’s Director of Applied Science. The Verge reported that Microsoft spokesperson Alex Haurek disputed those characterisations once they became public, saying the comments do not reflect the company’s position. Microsoft has not, according to the reporting, disputed that Hecht wrote them.
The primary document
The filing itself is available in full through the court record, and it rewards reading past the pull quotes. The “doom loop” and “theft of labor” lines are not asides buried in a footnote: they are the material the New York Times’ own lawyers selected as the most damaging evidence available, assembled specifically to support a request that the court rule against OpenAI and Microsoft without sending the case to a jury.
What to watch next
The case sits in the Southern District of New York, and the Times’ summary judgment motion is now before the court. A ruling in the Times’ favour would let the case proceed to trial on narrower grounds, or resolve some claims outright, without a jury needing to weigh the underlying facts. Given how much sealed material has already surfaced through this one filing, more of the record is likely to become public as the motion is briefed and argued.
