New York Times Says OpenAI Hid Evidence in ChatGPT Copyright Trial

The New York Times has accused OpenAI of concealing evidence in their long-running copyright battle, a claim that sharply escalates one of the most consequential legal fights over how AI companies use publishers' content to train and operate their models.

The allegation

In a filing this week, lawyers for the Times alleged that OpenAI withheld material relevant to how ChatGPT reproduces and summarizes the newspaper's articles, and that the company failed to preserve or produce records the court had ordered disclosed. OpenAI has denied wrongdoing in prior discovery disputes and has argued that its use of publicly available content falls under fair use.

The case, first filed in late 2023, has become the bellwether for a wave of publisher litigation against AI developers. Discovery battles have repeatedly spilled into public view, including earlier disputes over the deletion of training-data records and access to model outputs.

Why it matters

A finding that OpenAI concealed evidence could carry sanctions well beyond the usual discovery skirmish — courts have broad power to instruct juries to draw adverse inferences against parties that spoliate evidence. For a trial that will help define whether large-scale ingestion of copyrighted news content requires a license, that procedural posture could matter as much as the merits.

The stakes have grown with OpenAI's business. The company is reportedly preparing for an eventual IPO on the back of more than $25 billion in annualized revenue, and unresolved copyright liability remains one of the largest overhangs on that story.

The broader fight

Publishers in the US, Europe and Asia are watching closely. Licensing deals have proliferated — OpenAI, Google and Anthropic have all signed content agreements — but the Times case will set the reference point for what happens when negotiation fails.