New York Times accuses OpenAI of hiding evidence in ChatGPT copyright case

The New York Times and the Daily News have accused OpenAI of concealing evidence regarding its ability to search chat logs and training data for copyrighted material. This is the latest development in a two-year lawsuit alleging that OpenAI breached copyright by training AI models on the newspapers' content and reproducing it in ChatGPT outputs.

OpenAI had previously claimed it could not search its training corpus and that accessing chat logs was technically difficult and raised privacy concerns. However, in an April deposition, OpenAI data privacy engineer Vinnie Monaco allegedly revealed that the company had already conducted internal searches for copyrighted works and had built a database of about 78 million de-identified ChatGPT conversations. Additionally, OpenAI allegedly implemented a "Bloom" filter as part of "Project Giraffe" to track regurgitation of content shortly after the lawsuit was filed.

The plaintiffs had originally requested 120 million chat logs but negotiated down to 20 million, which were submitted last December. However, the sample was heavily redacted, making it "unusable" according to the court. The plaintiffs also claim OpenAI deleted billions of outputs after the suit was filed, violating a preservation order, and substituted millions of logs.

Lead counsel Ian B. Crosby stated: "If OpenAI genuinely believed that copying our clients' journalism was fair and legal, it wouldn't have hid the truth about having done it." The plaintiffs now seek sanctions, including preventing OpenAI from using the chat log sample as evidence, accepting as fact that logs would show major regurgitation, and covering legal fees.

OpenAI spokesperson Drew Pusateri denied the allegations, accusing the Times of trying to access private conversations as its case weakens. He reaffirmed OpenAI's commitment to user privacy and fair use.