New York Times-Led Media Coalition Asks Federal Judge to Sanction OpenAI for Hiding AI Training Data Evidence
The New York Times, the Daily News, and other US media outlets filed a motion on July 9 in Manhattan federal court asking a judge to impose sanctions on OpenAI for alleged evidence obstruction in their copyright infringement lawsuit, according to AP News. The filing alleges OpenAI “chose obstruction” over releasing datasets and ChatGPT conversation logs that could demonstrate how the company used millions of copyrighted news articles to train its AI models.
The plaintiffs are asking the court to penalize OpenAI for what they describe as “discovery misconduct” that could distort evidence in the case, Al Jazeera reported, citing a recent deposition of an OpenAI employee that allegedly contradicts the company’s earlier claims about its ability to search for copyrighted content in training datasets.
The Core Dispute
New York Daily News attorney Steven Lieberman said OpenAI has been “making misrepresentations” for two years about its capacity to search for copyrighted content in AI training data and logs. “This motion asks the court to punish OpenAI for hiding and destroying evidence showing how ChatGPT was trained on stolen journalism,” Lieberman said, according to Al Jazeera.
OpenAI has argued that releasing ChatGPT conversation logs would risk violating users’ privacy. Spokesperson Drew Pusateri called the allegations “blatantly false,” telling Reuters that “as the Times’ case weakens and they’ve been forced to drop claims against us, they’re persisting with their efforts to invade the privacy of people who have nothing to do with this case,” as reported by Al Jazeera.
The lawsuit dates to late 2023, when the New York Times sued OpenAI and Microsoft roughly a year after ChatGPT’s launch. At its center is whether AI chatbots are unfairly competing as information sources by siphoning web traffic without performing the journalistic work involved in gathering news, according to AP News.
Downstream Exposure for Agent Builders
The case is one of many brought by copyright holders, including authors, visual artists, and music labels, against OpenAI, Anthropic, and Meta Platforms for allegedly misusing their material to train AI systems. But for organizations building autonomous agents on frontier models, the unresolved question is not abstract.
Every AI agent deployed on GPT-5.6 Sol, Claude, or Llama inherits whatever legal exposure attaches to the model’s training data. If courts ultimately determine that training on copyrighted content without authorization constitutes infringement, the liability does not stop at the model provider. Enterprise customers and agent builders operating in regulated industries face the question of whether the outputs their agents generate, and the decisions those agents make based on copyrighted training data, create secondary liability.
The sanctions motion matters because it could force disclosure of exactly what content went into OpenAI’s training datasets. That disclosure would give agent builders their first concrete data point for assessing copyright risk in their own deployments, rather than relying on OpenAI’s assurances that its training practices are legally defensible.