OpenAI frames AI training as fair use, argues broader innovation benefits outweigh copyright claims
OpenAI publishes court filing excerpts defending generative AI training under fair use doctrine, positioning the legal dispute as a question of innovation versus copyright expansion.
Last verified:
OpenAI’s Fair Use Defense Takes Shape in Court Filings
OpenAI released excerpts from amicus briefs filed in the ongoing copyright litigation brought by The New York Times, positioning fair use doctrine as the legal linchpin protecting AI training from copyright restriction. According to OpenAI’s blog post, multiple stakeholder organizations have argued that restricting fair use for AI training would harm not only model developers but also medical researchers, educators, and people with disabilities who depend on AI-enabled tools and discovery methods.
The core argument OpenAI is advancing, as reflected in the briefs, centers on transformative use: that training AI systems on copyrighted material serves a fundamentally different purpose than the original works, does not provide substitute access to those works, and does not cause market harm actionable under the Copyright Act. OpenAI frames copyright law itself as designed to “promote creativity and progress, not to suppress lawful innovation or protect authors from competition.”
Who Joined OpenAI’s Legal Coalition
According to OpenAI, the American Council of the Blind, Association of Research Libraries, American Library Association, Consumer Technology Association, and Business Software Alliance filed joint briefs emphasizing that the stakes extend far beyond generative AI vendors. These organizations stress that restricting fair use training would affect downstream users and communities: medical researchers studying disease patterns, educators adapting materials for students with different learning needs, farmers applying agricultural knowledge through AI tools, and independent creators building new forms of artistic expression.
The Authors Alliance, a separate amicus filer, advanced a narrower economic argument: that copyright law has never restricted competition from new works that reduce the sales or visibility of older works, and that the Copyright Act deliberately leaves such market competition outside copyright holders’ control. This framing rejects what the brief terms a “market dilution theory” as beyond the Copyright Act’s intended scope.
Why This Matters
The fair use framing adopted by these amicus briefs represents a strategic pivot from arguing that AI training is economically beneficial to arguing that it is legally protected as transformative. If courts accept this logic, it would establish precedent that training on copyrighted material without license or direct compensation is permissible for sufficiently non-substitutive purposes. Conversely, if The New York Times prevails, publishers and content creators will gain a licensing leverage point over AI developers—potentially forcing negotiated agreements or shutting down training on recent copyrighted works entirely. The outcome will likely determine whether fair use remains a viable defense for future AI training or whether copyright licensing becomes a mandatory cost in the AI development pipeline.
Frequently Asked Questions
What does OpenAI claim about fair use and AI training?
OpenAI argues that generative AI training constitutes fair use because it is transformative, does not substitute for access to original works, and does not cause cognizable market harm under the Copyright Act.
Who filed amicus briefs in support of this position?
According to OpenAI, organizations including the American Council of the Blind, Association of Research Libraries, American Library Association, Consumer Technology Association, and Business Software Alliance filed briefs supporting fair use for AI training.
What is The New York Times' position in this case?
The New York Times lawsuit argues that AI training on copyrighted news articles without license or compensation violates copyright law—the opposite of OpenAI's fair use defense.