The US Supreme Court ruled 6-3 in June 2026 that generative AI companies can be held liable for copyright infringement when their training datasets include copyrighted works without authorization, rejecting the “transformative use” defense in that context.

The case, Authors Guild v. OpenMind AI Corp., was brought by a coalition of authors and publishers after their books appeared verbatim in outputs generated by a large language model trained without licensing their works.

The majority opinion, written by Justice Barrett, held that training an AI model on copyrighted text is not a transformative fair use because the purpose of the use is to replicate the information in the copyrighted work, not to comment on or transform it.

What the Court Decided

The majority rejected the argument that ingesting text for training is analogous to reading a book, noting that the scale, commercial purpose, and market substitution effects of AI training are qualitatively different from individual human reading.

The court did not rule on whether AI-generated outputs themselves infringe copyright when they do not reproduce specific protected expression. That question was explicitly left for lower courts to resolve in ongoing cases.

The ruling also did not address the legality of training on publicly available internet content, which involves different factual circumstances than the publisher-licensed books at issue in Authors Guild. That question remains unsettled.

What the Dissent Argued

Justice Sotomayor’s dissent, joined by Justices Kagan and Jackson, argued that the majority’s ruling will impede AI development by requiring license agreements for training data that do not exist at commercial scale.

The dissent argued that existing fair use doctrine under Section 107 of the Copyright Act is flexible enough to accommodate AI training as a transformative use, consistent with how courts have treated Google Books, search engine indexing, and other large-scale text analysis.

Industry groups representing AI companies filed amicus briefs supporting the dissent’s position. They argued the ruling creates an unworkable licensing requirement that will disproportionately benefit large companies that can negotiate bulk licenses over startups.

Immediate Industry Impact

Shares in publicly traded AI companies fell 3 to 8 percent on the day of the ruling. Several companies issued statements saying they are reviewing their training data practices and consulting with legal counsel.

OpenAI, Google, and Anthropic have all previously disclosed licensing agreements with some publishers and news organizations. Those agreements were entered before the ruling and may position them more favorably than competitors who did not license training data.

According to legal experts at the Stanford Center for Internet and Society, the ruling will likely accelerate licensing negotiations between AI companies and rights holders, with the Copyright Office expected to issue guidance on royalty frameworks within six months.

What Happens Next

The case returns to the Second Circuit to calculate damages. The court did not rule on the damages methodology, which is expected to be a major point of further litigation given the scale of training datasets.

Congress is expected to face renewed pressure to pass AI copyright legislation that provides clearer safe harbors. Several bills were introduced in 2025 but did not advance. The Supreme Court ruling makes inaction more costly for AI developers.

International implications are also significant. The EU’s AI Act includes text mining and data mining exceptions for research, and the EU Court of Justice may face similar questions under different legal frameworks in coming years.

Key Facts: Authors Guild v. OpenMind AI Corp.

Element	Detail
Court	US Supreme Court
Decision	6-3 for plaintiffs
Writing for majority	Justice Amy Coney Barrett
Key holding	AI training on copyrighted works is not fair use
Left undecided	Training on internet content; AI output infringement
Next step	Damages calculation in Second Circuit

Frequently Asked Questions

What does the Supreme Court AI copyright ruling mean for existing AI tools?

Existing AI tools are not immediately shut down or invalidated. The ruling creates legal liability for training on copyrighted works without a license, but ongoing use of already-trained models is a separate question. Companies may need to retrain future model versions on licensed datasets or pay damages for past training.

Does this ruling affect AI tools trained on public domain content?

No. The ruling applies specifically to copyrighted works. Public domain content, works with permissive licenses such as Creative Commons, and content where the owner has granted rights for AI training are not affected by this decision.

Can authors sue AI companies now?

The ruling establishes that the legal theory is viable, but individual authors still need to demonstrate that their specific copyrighted work was in a training dataset and that they suffered cognizable harm. Class action cases consolidating claims from many authors are the most likely practical legal vehicle going forward.