Publishers Seek Sanctions Against OpenAI Over Hidden Evidence

A group of U.S. publishers led by The New York Times filed a motion on July 9 in federal court in Manhattan seeking sanctions against OpenAI. According to the plaintiffs, the company deliberately obstructed access to training datasets and ChatGPT logs that could show how the system used millions of copyrighted news articles.

Allegations of hidden evidence

In a filing submitted Thursday, July 9, the publishers argue that OpenAI chose obstruction over turning over the datasets and ChatGPT conversation logs that could reveal how the system used copyrighted journalistic content. A lawyer representing the Daily News said that for two years the company made false statements about searching its own training data for copyrighted material.

OpenAI chose obstruction over turning over the datasets and ChatGPT logs that could reveal how it used protected news content - from the publishers' motion filed in federal court in New York

Scale of the dispute

Since The New York Times filed suit against OpenAI and Microsoft in late 2023, other publishers have joined the case, including the Chicago Tribune, several newspapers owned by MediaNews Group, the tech outlet Ziff Davis, and the investigative journalism organization Center for Investigative Reporting. The New York Times disclosed that its own legal costs tied to AI disputes have already topped $28 million, underscoring the scale of publishers' financial stake in the case.

At the heart of the dispute is the question of whether AI chatbots unfairly compete with media outlets as a source of information, capturing web traffic without doing the actual journalistic work of gathering news. That question goes beyond the matter of training data itself and touches the foundations of the business model underpinning both news publishers and the AI companies that rely on their content.

OpenAI's response

OpenAI rejected the allegations, calling them flatly false. The company attributes the limits on sharing ChatGPT logs to the need to protect user privacy, not any attempt to hide evidence. In its defense, it invokes the fair use doctrine, arguing that training AI models on content available online complies with U.S. copyright law.

The plaintiffs are asking the court to punish OpenAI for what is known as discovery misconduct, failures in the evidentiary process that could distort the record in a case seen as pivotal for the future of the news industry. The requested sanctions include covering the attorneys' fees publishers incurred while trying to obtain evidence they say was unlawfully withheld.

Stakes for the industry

The Times v. OpenAI case has long been viewed as a bellwether for the entire publishing industry, since its outcome could set the rules under which AI companies are allowed to train models on copyrighted content. For European and Polish publishers watching from the sidelines, what matters most is how the court weighs the fair use argument against commercial AI products generating billions of dollars in revenue.

The sanctions motion does not by itself resolve the underlying dispute, but it escalates tensions between the parties and could affect the pace of further proceedings. If the court grants the publishers' request, OpenAI may be forced to disclose a far broader set of training data than it has so far been willing to share, with consequences reaching beyond the New York case itself.

Sources: Washington Post (washingtonpost.com), The Hill (thehill.com), U.S. News (usnews.com), WRAL/AP (wral.com)