The New York Times Asks Court to Sanction OpenAI for Withholding Evidence

The New York Times, Daily News and other leading American publishers have asked a New York federal court to sanction OpenAI, accusing the company of withholding key evidence in a copyright infringement case.

The allegations against OpenAI

In the court filing, the publishers argue that the company "chose obstruction," deliberately impeding justice. Instead of handing over the datasets used to train ChatGPT, they say, OpenAI is keeping that information from the court. The publishers consider the data critical to proving their core argument: that OpenAI used their newspapers' copyrighted content without a license, permission or compensation for the authors.

• Main allegation: withholding the datasets used to train ChatGPT
• Plaintiffs: The New York Times, Daily News, Chicago Tribune, Las Vegas Review-Journal and many other publishers
• Court: the federal court for the Southern District of New York in Manhattan
• Core of the dispute: whether using copyrighted content to train AI models is lawful

What court sanctions are

Sanctions in civil proceedings are not just a fine. They punish contempt of court and violations of its orders. They serve as a means of legal coercion, forcing parties to comply with court rulings. If the judge finds that OpenAI deliberately withheld key evidence, the company faces monetary penalties, procedural restrictions (for example, a ban on using certain arguments in its defense) or an adverse inference, in which the court presumes the withheld materials contained information favoring the plaintiffs. Any of these outcomes would significantly reduce OpenAI's chances of winning.

The history of the copyright dispute and AI training

The New York Times filed the first lawsuit in December 2023, accusing OpenAI of using millions of articles from the newspaper's archive to train ChatGPT without the publisher's permission and without paying the authors. Dozens of other publishers soon joined the suit: New York Daily News, Chicago Tribune, Las Vegas Review-Journal, The Intercept and many other major newspapers and magazines.

It became one of the largest lawsuits in the AI industry's history. The central question remains open: may companies use published, copyright-protected content to train large language models? Is a publisher license and payment to authors required in that case? Regulators in the US, Europe and other countries are actively looking for an answer, but no single international standard exists yet.

Possible consequences

If the court sides with the publishers, it would require a rethink of the entire practice of training large language models. AI companies would have to either pay publishers for content, obtain explicit licenses before use, or exclude copyrighted content from training datasets entirely. That could radically change the economics of AI development, reduce the availability of quality training data and slow the development of new models.

What this means

The outcome of this case could become a global precedent. AI companies worldwide, including the developers of Claude, DeepSeek, Meta and other market leaders, will take a US federal court ruling into account. The impact will be especially noticeable in the US, where copyright carries the most weight. The outcome will affect startups' business models, the volume of investment in AI development and training-data strategy for years to come.

*Meta has been designated an extremist organization and banned in Russia.