Judge tosses most claims in authors' copyright fight with OpenAI

Authors threw the legal kitchen sink at OpenAI, but a federal judge just tossed most of it out the window.

U.S. District Judge Araceli Martínez-Olguín dismissed nearly all claims in the copyright lawsuits brought by comedian Sarah Silverman, Paul Tremblay, and other prominent writers against the ChatGPT maker.

The authors originally filed a sprawling complaint covering everything from vicarious copyright infringement and DMCA violations to negligence and unjust enrichment. In the end, almost none of it survived legal scrutiny.

The “every output is infringement” argument falls flat

The plaintiffs attempted to argue that because OpenAI ingested their copyrighted books to train its models, literally every single output generated by ChatGPT should be classified as an infringing derivative work.

Judge Martínez-Olguín firmly rejected that theory. The court noted that the authors failed to point to any specific output that copied their books or was substantially similar to their protected text. You cannot simply claim an output infringes just because ChatGPT can generate an accurate plot summary.

The DMCA claims met a similar fate. The authors accused OpenAI of stripping copyright management information from their works. But the judge pointed out that the writers failed to show how omitting this metadata knowingly induced or concealed infringement, nor did they prove OpenAI actually distributed copies of the books.

Claims of negligence and unjust enrichment were also thrown out. The court saw no legal duty of care or special relationship requiring OpenAI to safeguard public books, and called the authors' warnings of potential future financial harm purely speculative.

The only secondary claim left standing for now is a narrow prong of California’s unfair competition law regarding commercial use without permission. The authors have until March 13 to amend their complaint.

Why it matters

This ruling delivers an important reality check to copyright holders targeting generative AI. Courts are making it crystal clear that modern copyright law still requires concrete proof of substantial similarity in outputs, not sweeping legal theories about derivative works.

For AI builders, this is a major tactical sigh of relief. It knocks down high-risk claims that carry statutory damages, like the DMCA violations, and prevents plaintiffs from claiming that every single query run through a chatbot constitutes copyright theft.

However, OpenAI is far from out of the woods. The company deliberately avoided asking to dismiss the core direct infringement claim, choosing instead to save its powder for a summary judgment battle on fair use.

We think this ruling effectively clears away the legal clutter to set up the only fight that actually matters: whether ingesting copyrighted works for model training qualifies as fair use under federal law. With parallel lawsuits from George R.R. Martin and The New York Times looming, the true showdown is still ahead.