Four publishing-industry organisations have asked a federal court in New York to rule that OpenAI and Microsoft cannot rely on fair use for training AI models on copyrighted books. The Association of American Publishers (AAP), the International Association of Scientific, Technical & Medical Publishers (STM), the Independent Book Publishers Association (IBPA) and Copyright Clearance Center (CCC) filed a joint amicus brief on 29 October in In re OpenAI, Inc. Copyright Infringement Litigation, AAP said.
An amicus brief is an argument from outsiders to a case, not a decision. The four groups are not parties, the court has not ruled, and the claims in the brief are allegations drawn from the plaintiffs’ evidence.
What the brief asks for
The 25-page brief appears on the docket of the consolidated case, No. 25-md-3143 before the US District Court for the Southern District of New York, as an attachment to docket entry 2304. It supports the summary-judgment motions of the class plaintiffs, led by authors, and of the news plaintiffs. AAP says the four groups want the court to grant the plaintiffs’ motion and deny the defendants’.
The consolidated case gathers suits filed over the past three years against OpenAI, and against Microsoft for what the plaintiffs say was its knowing support of OpenAI. One is the September 2023 action by the Authors Guild and novelists including David Baldacci, John Grisham and Scott Turow. TSN has covered USA Today Co.’s suit against OpenAI, which asked to be treated as related to the same proceedings.
The brief’s central legal claim is that AI is no special case. “Using copyrighted works to train generative AI models does not warrant special treatment,” it says. Fair use, it argues, “is intended to adapt—not yield—to changes in technology”, and “The technological innovation of AI is no talisman for fair use.”
The evidence it leans on
The brief says the copying is not transformative, because large language models use books for their expressive content and for the same purpose as the originals. It then cites material from the plaintiffs’ statement of facts. It says ChatGPT has produced sequels that substitute for George R. R. Martin’s unfinished The Winds of Winter, and that OpenAI pursued a “research mission” to build a model that can “write the last two books of . . . A Song of Ice And Fire.”
It also quotes the defendants’ own staff. According to the brief, an OpenAI researcher wrote of a goal to “make the most artisanal slop”, and a Microsoft employee concluded that the harm from these models included “authors . . . who were no longer able to sell their books because there are so many substitutive AI-generated books out there.” Another OpenAI researcher is quoted describing a “doom loop” for the company’s models and the web. These are the brief’s characterisations of evidence that the court has not yet weighed.
The second argument is about money. The brief says a market for licensing works for AI training already exists and is growing, pointing to deals across the book trade and to a collective licence offered by CCC. It cites a Vantage Market Research estimate that this market, across all content types, was worth $5.2 billion in 2025 and could reach $55.8 billion by 2035. Those are the brief’s figures and its source’s projection, not court findings.
AAP chief executive Maria Pallante said the brief explains “why the failure of AI companies to pay for the works they need is an unjustified and damaging departure from the operation and integrity of the law.”
A crowded docket, and a decision still to come
Judge Sidney Stein invited would-be amici to seek leave and attach their proposed briefs, with a deadline of 30 October and a 6,000-word cap, according to the Center for AI and Digital Policy’s case page. The AAP brief certifies 5,998 words. Publishers Weekly and The Bookseller also reported the filing.
The US government has filed on the other side. As TSN reported in its analysis of the Third Circuit’s ROSS ruling, the Justice Department told this court on 1 September that training on copyrighted material is not, in itself, infringement. That was a statement of interest, not a ruling either.
Why the licensing-market argument carries the most weight
This is TSN’s analysis. Of the brief’s two lines of attack, the market argument is likely to matter more. Courts weigh harm to the market for the original work heavily, and the brief is trying to show that a licensing market is real, priced and growing, not hypothetical. If the judge accepts that, unlicensed training looks less like a new use and more like skipping a bill.
The substitution evidence is more striking but harder to carry. Quotes about “artisanal slop” and sequels to Martin’s novels make vivid copy. Yet a court will ask what the models actually produce and how often, and those questions turn on the full evidence, not on selected lines in an amicus filing.
The filing also lands in a divided landscape. The Third Circuit has rejected fair use for a non-generative legal research tool, while Anthropic is seeking a fair-use ruling for Claude. AAP expects Judge Stein’s decision to be the first opinion from a New York court on copyright and AI training. That makes his reasoning on market harm the part to watch, whichever side wins.
More amicus filings are due by 30 October. TSN tracks the wider fight on its AI copyright lawsuits hub.
Related
- AI Copyright Lawsuits: Cases, Rulings and Deals
- USA Today and 13 Sister Companies Sue OpenAI: What the Complaint Alleges
- Not Fair Use: What the Third Circuit Just Said About Training AI on Westlaw Headnotes
- Anthropic Seeks Fair-Use Ruling on Claude Lyrics as Judge Weighs First AI Trial
Related stories
- USA Today and 13 Sister Companies Sue OpenAI: What the Complaint Alleges
- Not Fair Use: What the Third Circuit Just Said About Training AI on Westlaw Headnotes
- Anthropic Seeks Fair-Use Ruling on Claude Lyrics as Judge Weighs First AI Trial
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
