The decision mirrors a ruling that nixed some core claims from a similar suit brought against Meta and its AI tools
A group of authors, including Sarah Silverman and Ta-Nehisi Coates, were dealt an arguably expected setback in a copyright infringement case against OpenAI. On Monday, Feb. 12, a federal judge threw out some, but not all of the claims contained in the lawsuit, which alleges OpenAI’s ChatGPT program was trained on copyrighted books.
The ruling mirrors a decision in a separate case brought by some of the same author plaintiffs (including Silverman) against Meta over how it trains its artificial intelligence language models. Like Meta, OpenAI had asked a judge to dismiss some of the authors’ broadest claims about the company’s liability, and was largely successful in that endeavor.
The big claim the judge dismissed was the vicarious copyright infringement allegation, which essentially argued that every answer generated by ChatGPT should be considered infringing because the language model was allegedly trained on unlicensed, copyrighted material. The judge called this claim “insufficient,” saying the plaintiffs “fail to explain what the outputs entail or allege that any particular output is substantially similar — or similar at all — to their books.”
Furthermore, the judge said there was no evidence to support the authors’ claim that OpenAI had violated the Digital Millennium Copyright Act by removing pertinent copyright information from the material it trained its model on as a way to hide its alleged infringement. And negligence and unjust enrichment claims were also tossed.
The lawsuit still has some teeth, though. The judge declined to dismiss a claim of unfair competition that’s rooted in the core allegation that OpenAI used copyrighted works to train its language models for commercial profit.
Lawyers for OpenAI and the author plaintiffs did not immediately return Rolling Stone’s requests for comment.
