OpenAI highlights 20 filings backing fair use for AI training as copyright ruling awaits

OpenAI put a public spotlight on its copyright defense on October 30, publishing excerpts from 20 filings supporting fair use for AI training. The collection brings together librarians, authors, technology groups and the U.S. government. But it is advocacy, not a ruling: a New York federal judge is still considering both sides’ requests for judgment.

The consolidated litigation combines lawsuits against OpenAI and Microsoft brought by The New York Times, other news publishers, Ziff Davis and authors led by the Authors Guild. The parties have asked Judge Sidney H. Stein in the Southern District of New York to resolve fair-use questions through summary judgment—decisions based on the case record rather than a trial.

Before the showcase, competing filings

The latest round of outside arguments arrived around an October 30 deadline for amicus briefs, submissions from interested parties outside the lawsuit. On October 29, Public Knowledge filed in support of OpenAI’s fair-use motion. Its central request was to evaluate training separately from possible infringement in acquiring data or producing AI outputs.

Meredith Rose, the group’s senior policy counsel, argued that training extracts information such as word frequency from works. She acknowledged risks to creators from unchecked AI adoption but said weakening fair use was the wrong response. Her position distinguishes the act of learning patterns from questions about copying at other stages.

That same day, the Association of American Publishers, two other publishing associations and Copyright Clearance Center jointly filed for the plaintiffs. Their counterargument centers on licensing: they say AI companies should pay for the works they use, and that uncompensated training transfers value away from those who financed and created the material.

The publisher groups contend that a market for AI training licenses already exists and is growing. In their view, payment sustains human authorship while giving AI developers reliable access to useful content. They reject the idea that a new technology should displace the system of exclusive rights and negotiated exchange established by copyright law.

The fair use doctrine is intended to adapt—not yield—to changes in technology.

Joint brief from AAP, STM, IBPA and Copyright Clearance Center

OpenAI’s coalition has different boundaries

OpenAI’s October 30 post, titled “Fair use benefits everyone,” highlighted the other side of that argument. Its selections include library and disability organizations, legal scholars, computer scientists, creators and Khosla Ventures. Several excerpts frame training as transformative: using works for a different purpose, rather than offering them to the public to read.

The Computer and Communications Industry Association’s excerpt describes copyrighted works as material analyzed for statistical patterns, enabling a tool to generate new content at a user’s direction. A joint submission from library, accessibility and technology groups emphasizes people who use models built by others, including open-source systems, rather than training models themselves.

Authors Alliance supports the defendants’ motions on a narrower question: “market dilution,” meaning competition from new AI-generated works even when they do not copy protected expression. It argues that copyright should not shield an author from other works sharing a genre, facts or audience. The group expressly takes no position on data acquisition or whether particular outputs infringe.

That distinction also reflects concern for authors who build on earlier work. Authors Alliance warns that broad market-dilution claims could burden scholars, biographers and smaller publishers. It argues that multiplying the number of lawful competitors does not turn ordinary competition into a copyright injury, even if the economic pressure is real.

Nor does every selection endorse a winner. IPWatchdog reports that law professor Matthew Sag’s brief backs neither side, although OpenAI included it. Sag views training as transformative under ordinary conditions, but also asks the court to consider books obtained from “shadow libraries” when assessing market effects, and systems designed to supplant a particular source.

The next move belongs to the court

The collection also includes an older government intervention. In a September 1 Statement of Interest, the Justice Department urged a fair-use finding and warned that mandatory licensing could favor technology firms able to pay, harming U.S. competitiveness and national security. That is the government’s argument, not a new court determination.

Stein must now weigh the competing requests on the record before him. The parties’ reply briefs are due November 20. OpenAI’s publication showcases arguments for protecting training; the pending motions will determine whether those arguments persuade the judge in this litigation.