OpenAI Accused of Hiding Billions of Logs and Lying About Search Capabilities in NYT Copyright Lawsuit

Summary

OpenAI faces calls for "serious sanctions" after news organizations including The New York Times alleged the AI company repeatedly lied for years about its technical ability to search ChatGPT training logs. According to court filings, OpenAI claimed for two years that searching large samples of logs was technically infeasible and burdensome, but testimony revealed the company had already conducted such searches and possessed two large de-identified samples (10 million and 78 million logs) that could have been made available during discovery.

The evidence at stake is crucial to both sides of the copyright infringement case. News organizations claim OpenAI trained ChatGPT on millions of copyrighted articles and that the chatbot regurgitates their content, violating copyright protections. OpenAI argues its use constitutes transformative fair use. The alleged concealment of searchable logs has frustrated plaintiffs' efforts to prove infringement and extended the already lengthy discovery process.

OpenAI's privacy engineer Vincent Monaco allegedly revealed during an April deposition that the company had previously searched the logs specifically for New York Times content as part of research into creating a filter to block regurgitation of copyrighted material. News organizations argue this admission exposes OpenAI's deliberate misrepresentation to the court and contends the company obscured evidence to shield its fair use defense. OpenAI disputed these characterizations, claiming the lawsuit was weakened and accusing news organizations of attempting to invade user privacy.

• OpenAI had already searched the logs for New York Times content as part of developing a content filter, directly contradicting its claims about technical limitations