The copyright litigation over how OpenAI trained ChatGPT has reached its most consequential stage so far, and outside groups are crowding into the courtroom. On October 30 OpenAI published a post titled "Fair use benefits everyone" that collects excerpts from amicus briefs supporting its position in the consolidated cases brought by The New York Times, other news organizations and book authors. Publisher groups filed briefs on the other side the same week.
The cases are grouped as In re OpenAI, Inc. Copyright Infringement Litigation in federal court in Manhattan (the Southern District of New York), where Judge Sidney Stein presides and Microsoft is also a defendant. According to Authors Alliance, both sides asked for summary judgment in cross-motions this autumn, oppositions arrived on October 23, and amicus briefs were due around October 29. The News/Media Alliance describes this as the first time a U.S. court will decide summary judgment on OpenAI's conduct and on real-time use of news content in AI answers.
OpenAI's selection is broad. It quotes filings from library and accessibility groups (the American Council of the Blind, the Association of Research Libraries and the American Library Association, joined by the Consumer Technology Association and the Business Software Alliance), Public Knowledge, Engine and the Developers Alliance, the Foundation for American Innovation, writers such as Kevin Kelly, and law professors including Matthew Sag. It also cites the United States government. Reuters reported in September that the administration's statement of interest, filed on September 1, appeared to be the first time Washington had weighed in on the wave of AI training cases; such filings are advisory. OpenAI's own framing is combative: it says the Times wants to discard decades of fair-use precedent in pursuit of what it calls an undeserved payday.
Some briefs target narrow legal theories. Authors Alliance, which supports the defendants, focuses only on "market dilution", the idea that a model harms authors by enabling floods of new, non-infringing works that compete with theirs. The group traces the theory to Judge Vince Chhabria's 2025 ruling in Kadrey v. Meta, points out that Judge William Alsup turned down a comparable argument in Bartz v. Anthropic, and says the class plaintiffs have asked Judge Stein to adopt it. Its position is that competition from new works has never been a copyright injury, however large the volume.
Publishers see the case very differently. The News/Media Alliance filed a brief on October 30 with three cosigners, ProPublica, the National Press Photographers Association and the National Newspaper Association, arguing that training on news and using it to power retrieval-based answers is substitutive rather than transformative. The alliance cites research it says shows ChatGPT refers 96% fewer visitors to news sites than classic Google search does, and warns that a fair-use finding would undercut a growing market for licensing news to AI companies. Its chief executive, Danielle Coffey, called the case existential for the industry.
The Wikimedia Foundation took neither side. As reported by The Verge, its brief argues that scraping and pre-training on copyrighted works can be fair use when the result is something new, but that systems which retrieve copies in response to user queries and deliver them in place of the originals should need a license. For Wikipedia, that would mean honoring Creative Commons attribution and share-alike terms. General counsel Stephen LaPorte warned that a ruling which does not apply fair use carefully could threaten free knowledge and open-source development alike.
None of these filings decides anything. Amicus briefs inform the court but carry no binding weight, and the excerpts OpenAI chose are, by design, the supportive ones. The outcome will turn on Judge Stein's reading of the four fair-use factors on the record the parties built in discovery, and on whether he treats training and output-time retrieval as one use or two. That distinction, raised from different directions by Wikimedia and the publishers, could matter well beyond this case.