Court wants OpenAI to share 20 million ChatGPT conversations; company responds with a ‘warning’

OpenAI is pushing back against a federal court order that would force it to turn over 20 million ChatGPT conversation logs to the New York Times in a copyright infringement battle. In a court filing, the artificial intelligence company called the demand a sweeping invasion of user privacy that goes far beyond what's needed to resolve the lawsuit.

The vast majority of these logs—more than 99.99%—bear no connection to the copyright dispute, OpenAI argues. The company warns that complying would expose intimate exchanges from millions of users caught in the crossfire of a legal fight that has nothing to do with them. The Times sued OpenAI for allegedly using its articles without permission to train ChatGPT.The logs aren't isolated prompts and responses—they're entire conversations with dozens of back-and-forth exchanges, randomly sampled from two years of ChatGPT use.

OpenAI's security chief Dane Stuckey warns this means private discussions about sensitive topics could land in the hands of the Times' lawyers and outside consultants.OpenAI's security chief Dane Stuckey warns this means private discussions about sensitive topics could land in the hands of the Times' lawyers and outside consultants.Magistrate Judge Ona Wang sided with the Times last week, setting a November 14 deadline for OpenAI to produce the anonymized chats.

She ruled that privacy safeguards and data scrubbing offer sufficient protection. OpenAI had proposed narrower alternatives, like searching only for chats mentioning Times articles, but the newspaper refused.

NYT dismisses privacy claims

The Times isn't buying OpenAI's privacy argument. A spokesperson dismissed the company's concerns as deliberate misinformation, pointing out that OpenAI's own terms allow it to use chats for legal purposes.

The newspaper insists user privacy remains protected through anonymization and legal guardrails.OpenAI counters that no court has ever ordered such a massive dump of private data. The company likens it to forcing Google to hand over millions of Gmail messages in any lawsuit—a comparison meant to highlight what it sees as dangerous legal territory.The clash is part of a broader wave of litigation testing whether AI companies can freely use copyrighted work as training material. OpenAI says it's exploring every avenue to block the data handover, while developing encryption features that would keep future chats private even from OpenAI itself.