A New York Times-led group of news publishers is asking a federal judge in Manhattan to sanction OpenAI, escalating a copyright fight over how ChatGPT and other AI systems were built.
The publishers say OpenAI has withheld or destroyed evidence that could show whether copyrighted journalism appeared in training data, datasets, or ChatGPT logs. OpenAI denies the accusations and says the demands would intrude on user privacy while the company defends its use of public material as fair use.
The dispute matters because it is no longer only about whether AI companies copied articles. It is also about what evidence courts can demand from AI developers, how much private user data can be searched in discovery, and whether publishers can prove that AI tools compete with their reporting. The answer could affect publishers, software makers and ordinary ChatGPT users whose logs may become relevant in discovery fights.
What changed
The Associated Press reported that the publishers asked the court to penalize OpenAI for alleged discovery misconduct after a recent deposition. The plaintiffs include The New York Times, the Daily News and other media organizations that argue OpenAI and Microsoft used millions of news articles without permission.
TechCrunch separately reported that the publishers say OpenAI had internal tools and datasets that could search for copyrighted journalism, even as the company resisted producing similar evidence in the case. CourtListener docket metadata for the consolidated OpenAI copyright litigation shows recent filings as of July 10, 2026.
OpenAI has framed the fight differently. In prior public updates about the Times case, the company said broad data-retention demands conflict with its privacy commitments and argued that AI training on publicly available internet material is protected by fair use. Its latest statement to AP rejected the sanctions allegations and said the company would keep defending user privacy.
Why it matters
If the judge grants sanctions, the ruling could shape the evidence available in one of the most important AI copyright cases in the United States. Sanctions can range from fee awards to findings that make it harder for a party to dispute certain facts later in the case.
For publishers, the case is about compensation, attribution and whether AI products can answer readers without sending traffic back to original reporting. For AI companies, it is about whether training models on large bodies of online text can remain legally viable without negotiating licenses for every source.
The privacy issue is also practical for users. Courts may need to decide how to balance publisher discovery requests against logs that can include personal, confidential or business-sensitive prompts from people who are not parties to the lawsuit.
What to watch next
The next key step is how the federal judge handles the sanctions request and any response from OpenAI and Microsoft. A narrow ruling could focus only on discovery conduct. A broader one could influence how future AI cases handle training data, user logs and publisher evidence.
The underlying copyright question remains unresolved: whether training large AI models on journalism is fair use, infringement, or something courts and companies will eventually push toward licensing deals.
