A diverse group of stakeholders — including authors, journalists, creators, academics, accessibility advocates, legal experts, librarians, economists and business leaders — have filed amicus briefs arguing that the longstanding doctrine of fair use should apply to the use of copyrighted works to train generative AI models. The filings present legal, practical and policy reasons for upholding fair use in this context.
Those submissions contend that fair use has long fostered innovation that produces public benefits, and that the New York Times’ lawsuit seeks to overturn decades of legal precedent by narrowing fair use in service of a commercial claim.
Key legal and policy arguments
- The briefs say AI training serves a fundamentally different purpose than the original works and does not provide a substitute for public access to those works. Under this view, training models is transformative and therefore favors fair use.
- Several amici urge the court to reject plaintiffs’ attempt to expand copyright protection to cover “market dilution” from new works that merely compete in subject matter or audience. They argue aggregate competitive pressure from noninfringing outputs should not qualify as cognizable market harm under 17 U.S.C. § 107(4).
- The filings emphasize that copyright exists to promote creativity and progress, not to suppress lawful innovation or to shield rightsholders from ordinary market competition.
Who filed and what they said
Organizations such as the American Council of the Blind, Association of Research Libraries, American Library Association, Consumer Technology Association and Business Software Alliance filed a joint brief stressing that fair use enables researchers, educators, people with disabilities, farmers, independent creators and other communities to study broadly and to use tools that make that knowledge practical. They warned that restricting fair use would harm both institutions that build AI and the people who rely on it.
Startups and technology advocates argue that training foundation models aims to build adaptable, general‑purpose tools rather than reproduce training works. The briefs state that the outputs startups pursue generally bear no resemblance to the purposes of the works in their training datasets.
One amicus notes that the benefits of AI training — including new knowledge, software and analytic capabilities — should factor into assessments of market effects under fair use, and suggests consideration of national‑security values tied to those capabilities.
Firsthand accounts from creators and practitioners
Independent creators Sarah Rose Siskind and Michael Wall describe how AI tools assist their businesses without displacing creativity. According to their briefs, AI reduces administrative time and cost, helps attract clients, and allows them to devote more time to composing and creating. Both say they doubt they could operate their businesses in the same form without AI.
Other amici point to precedent where fair use protected information‑access tools (for example image search and full‑text book search), libraries’ accessible digital services for readers with print disabilities, and the free publication of legal standards — arguing that fair use can similarly ensure AI’s benefits remain widely accessible rather than concentrated.
Press and academic perspectives on market effects
A brief filed by a Pulitzer Prize‑winning journalist and a law professor argues that OpenAI’s training on newspaper articles did not harm the reasonable commercial expectations of news plaintiffs: publishers retain their exclusive rights to sell access to articles, and those rights were not usurped by the model training. The brief warns that narrowing fair use to exclude training would impede scientific and creative progress and harm economic mobility.
Why the court’s ruling matters
Amici collectively argue that a ruling for the plaintiffs would extend copyright protection beyond the limited monopoly Congress intended, potentially stifling technologies capable of extraordinary information dissemination and new forms of thought and expression. Preserving fair use for AI training, they contend, balances the incentives to create original works with the societal benefits of tools that learn from those works.
The full amici filings are on the public court docket and set out detailed legal and factual accounts from many sectors to support the conclusion that training large language models on copyrighted materials is fair use under 17 U.S.C. § 107.
