OpenAI posted excerpts from court briefs on its website on October 30, 2026, arguing that fair use doctrine protects training AI on copyrighted works S¹. Fair use is the legal principle that permits use of copyrighted material without permission in certain circumstances. No court has ruled that AI training qualifies as fair use. The briefs OpenAI chose to highlight represent one side of an unresolved legal fight. OpenAI has not released the full text of any brief, only excerpts it selected, and the article names no specific court cases, jurisdictions, or individual filers, so the original briefs cannot be checked against OpenAI's selected passages — the claim that these briefs support fair use for AI training rests on a vendor's own curation, with no independently verified case names, no released source documents, and no court rulings to confirm it S¹.
My read: This is the clearest public advocacy push I've seen from OpenAI on fair use. The company is curating sympathetic voices — creators, academics, accessibility advocates, to frame fair use as settled doctrine applied to a new technology. It isn't settled. What I'd watch: whether any of the excerpts name specific cases, jurisdictions, or judges, because the article as published doesn't, and that makes independent verification impossible. The headline, "fair use benefits everyone," is a lobbying position, not a legal conclusion.
The article pulls selected passages from court briefs filed by creators, academics, accessibility advocates, and legal experts S¹. OpenAI says these briefs show how fair use supports AI training and innovation S¹. The company does not publish the full text of any brief, only excerpts it chose.
Google News indexed the article the same day it appeared S², but no independent legal reporting has corroborated the contents or the identities of the brief authors.
Fair use framing extends a years-long platform-publisher dispute
OpenAI has been building toward this argument for months. In January 2024, the company published a separate page on its relationship with journalism, stating it wanted to build AI tools that give people more capability while recognizing friction with news publishers P⁶. That page framed licensing deals as part of the solution but did not concede that training on copyrighted news articles required permission. The company has also been studying how people actually use its tools in workplaces, part of a broader effort to position itself as a toolmaker rather than a content competitor.
Google, which faces its own legal pressure over how it handles news content, has been making parallel arguments since at least 2020. Richard Gingras, then VP of News at Google, wrote in June 2020 that the decades-old news business model was under severe strain and cast Google's role as sending traffic to publishers P³. The same tension — platforms using news content to build products while publishers demand compensation, now runs through the AI training debate.
For a news publisher, a law firm advising media clients, or a university technology transfer office, the practical question is unchanged. Will a court treat AI training like the search-engine indexing that Google successfully defended, or like something new that requires licences? A media-company general counsel who today assumes training requires a licence would need to prepare for the possibility that a court disagrees, while a startup building a model on scraped text would face the inverse risk — that a ruling finds training is not fair use and forces a halt or a licensing round. OpenAI's article is an argument that it is the former. No ruling has confirmed that.
The next checkpoint is not a blog post. It is whatever court, in whatever jurisdiction, first issues a decision on whether training a language model on copyrighted text without a licence is fair use. OpenAI's article does not name that case, and the company has not said when it expects one.
Sources: S1 — Fair use benefits everyone · S2 — Fair use benefits everyone - OpenAI · P3 — Setting the record straight on news · P4 — lewisdonovan/google-news-scraper · P5 — fairlearn/fairlearn · P6 — OpenAI and journalism | OpenAI
Written from 6 sourced items, 5 of them primary.
