OpenAI claims the Times cheated to get ChatGPT to regurgitate articles

OpenAI has claimed in a motion filed Monday that The New York Times used “deceptive prompts” to get ChatGPT to regurgitate its content. For that and other reasons, the company is asking the US District Court in southern New York to dismiss several of the claims in the outlet’s copyright infringement lawsuit.

OpenAI asserts that the Times exploited a bug that it’s currently working to fix and that the outlet fed articles directly to the chatbot to get it to spit out verbatim passages. “Normal people do not use OpenAI’s products this way,” the company says, citing an April 2023 Times article titled “35 Ways Real People Are Using A.I. Right Now.” This is all very similar to the arguments OpenAI made in its public response in January.

Times lead counsel Ian Crosby told The Verge in an email that calling the outlet’s efforts a hack is a mischaracterization and that the outlet was “simply using OpenAI’s products to look for evidence that they stole and reproduced The Times’s copyrighted works.” He added that OpenAI doesn’t deny “it copied Times works without permission within the statute of limitations.”

The Times sued OpenAI and Microsoft in December, claiming the companies trained their AI models on its content and that their chatbots could reproduce the stories verbatim. The publication alleged that this deprives it of revenue and compromises its relationship with its readers.

OpenAI is looking to partially dismiss the Times’ count of direct copyright infringement “to the extent it is based on acts of reproduction that occurred more than three years before this action.” It also asks the court to dismiss other allegations: that OpenAI contributed to the infringement; that it had failed to remove infringing information; and that it created unfair competition by misappropriation. The Times lawsuit also alleges counts of trademark dilution, common law unfair competition by misappropriation, and a vicarious copyright infringement claim.

OpenAI similarly whittled down complaints in a lawsuit from Sarah Silverman and other authors to a single direct copyright infringement claim. As successful as its bid was and this one could be, the two aren’t the only lawsuits against AI companies. Startups like OpenAI, Anthropic, and Stability AI are looking into a steadily widening maw of legal action right now, some of it from experienced and litigious organizations with sometimes decades of copyright battles under their belts. As The Verge’s Nilay Patel and Sarah Jeong recently discussed on the Decoder podcast, the cases have the potential to upend or even obliterate the nascent industry.