Summary
On September 17, 2026, the news plaintiffs in the OpenAI copyright multidistrict litigation filed a public version of their motion for summary judgment, removing redactions that no party had moved to keep sealed. The 92-page brief, filed for The New York Times, the Daily News plaintiffs, the Center for Investigative Reporting, The Intercept and Ziff Davis, quotes internal Microsoft documents describing large-model training as "an astonishing theft of unprecedented proportions" and "the largest theft of labor in human history," and quotes OpenAI's head of ChatGPT writing that publishers face an "existential threat" from products that are "largely substitutive." It asks Judge Sidney H. Stein to rule for the plaintiffs on infringement at each stage of the AI pipeline, on fair use, on removal of copyright management information under the DMCA, and on per-article statutory damages. The brief quotes the defendants' own documents and testimony; the exhibits behind most of the quotations remain sealed.
What Happened
The brief was filed as an attachment to a one-page letter from Davida Brook of Susman Godfrey, the news plaintiffs' liaison counsel, docketed as ECF No. 1977 in No. 25-md-3143. The letter says the plaintiffs removed redactions under paragraph 9.c of the court's September 3 stipulated omnibus sealing order (Dkt. 1685) for material that no party or third party had sought to seal. OpenAI and Microsoft filed revised public versions of their own summary-judgment briefs against the news and book plaintiffs the same day, and over September 17 and 18 the news plaintiffs filed seven public batches of exhibits while other parties filed dozens of sealed declarations.
The introduction attributes the two "theft" quotations to Microsoft's Director of Applied Science. TechCrunch identified him as Brent Hecht and reported that the quotations come from a January 2023 internal memo. The fuller passage quoted in the brief reads: "millions of people around the world will soon consider large models 'hoovering up' all their work to be 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." A separate Microsoft document quoted in the brief states: "Our AI content strategy has started a 'doom loop' that will hurt the performance of our models and the entire web at the same time," and the brief says Microsoft's own data recorded 83% to 93% drops in click-through rates to The Times's and the Daily News plaintiffs' domains, and 51% to 94% for Ziff Davis domains, when users were served Copilot answers instead of traditional Bing results.
On acquisition, the brief states that OpenAI's mid-training datasets contain more than 91,692 copies of the plaintiffs' works and that a post-filtered Common Crawl dataset listed 2,064,805 documents from nytimes.com and 2,182,079 from chicagotribune.com. It describes two data transfers between the companies: "Project Taxi," under which Microsoft provided OpenAI with its Bing index, and "Project Mango," in which Microsoft helped OpenAI "collect as many of the documents as possible" for training, yielding a dataset with at least 160,903 unique works from the plaintiffs. It quotes OpenAI researcher Nick Ryder telling Greg Brockman about "a hack to get around nytimes paywall," and Brockman replying "ah nice," and it alleges that OpenAI systematically removed copyright management information from articles before training.
The brief also quotes Brockman writing that OpenAI's models were "excellent at news," and Microsoft chief executive Satya Nadella testifying in a deposition that "anything that is paywalled should be licensed by anyone who wants to use it" for grounding or training, and that had he known OpenAI trained on paywalled material he would have invoked Microsoft's right to require OpenAI to retrain its models. TechCrunch and The Washington Post reported the unsealed quotations the same day, and TechCrunch noted that the quotations come from the plaintiffs' brief rather than from the sealed exhibits themselves. OpenAI and Microsoft did not respond to TechCrunch's requests for comment.
Litigation over the sealing continued after the filing. On September 21, Magistrate Judge Ona T. Wang ordered the parties to propose a briefing schedule on a motion by Bloomberg Industry Group to intervene for the limited purpose of opposing sealing and redactions, and counsel for Microsoft and OpenAI filed a joint letter motion seeking more time to oppose the news plaintiffs' motion.
Why It Matters
The filing puts into the public record the defendants' own internal descriptions of the practices at issue, less than three weeks after the Justice Department filed a statement of interest supporting OpenAI's fair-use defense in the same proceeding. Several quotations speak to the fourth fair-use factor, the effect of the use on the market for the original works; the plaintiffs use them to argue that ChatGPT and Copilot substitute for news rather than transform it.
The quotations are the plaintiffs' selection and characterization, and the defendants' summary-judgment briefs and oppositions will contest their context and meaning. The "astonishing theft" passage describes how the public would come to view large-model training, which the plaintiffs present as an admission; whether the court reads it that way is not known. Judge Stein has not ruled on any of the cross-motions, and the outcome of the motion to intervene will determine how much more of the underlying record becomes public.
§ How to read the metadata
- Landmark
- Fundamentally alters the trajectory; 2–5 per year.
- Major
- Meaningfully shifts the landscape; 2–4 per month.
- Notable
- Worth documenting; significance can be upgraded later.
- Confidence
- High = primary sources corroborate. Medium = credible secondary only. Low = provisional. Disputed = credible sources disagree.
- Contestation
- Uncontested = no formal challenge. Contested = at least one challenge open. Superseded = replaced by a later entry. Unresolved = dispute still open.