The New York Times and The Daily News claim that OpenAI has been lying about its ability to search customer chat log data and training data sets for its copyrighted works. It’s the latest escalation in a two-year lawsuit against the AI ​​firm for allegedly violating copyright law by training its generative AI models on Times content and reproducing that journalism in user results.

Throughout the case, OpenAI has argued that it lacked the ability to search its own training corpus. It also argued that searching or producing its massive collection of ChatGPT conversations would be technically cumbersome and raise user privacy concerns because the records would have to be retrieved, processed, and anonymized. The outlets searched that data to determine whether their copyrighted journalism was present in the OpenAI training data set and whether and how often ChatGPT generated responses using or playing their content.

In a court-ordered deposition in April, OpenAI data privacy engineer Vinnie Monaco reportedly revealed that OpenAI had already conducted internal searches and evaluations of its training corpus to search for copyrighted journalistic works.

Monaco’s statement also allegedly revealed that, since before the NYT filed its lawsuit, OpenAI had already amassed a database of around 78 million unidentified ChatGPT conversations that it was using internally to determine the extent to which it was infringing on the works of others. In addition to that data set, OpenAI also allegedly implemented a “Bloom” filter as part of a toolset called “Project Giraffe,” which detected and kept track of regurgitation in the results, shortly after the lawsuit was filed.

Those last two revelations are particularly significant. The plaintiffs had originally asked OpenAI to provide a sample of 120 million chat records, but OpenAI had negotiated to reduce the sample to just 20 million. OpenAI finally submitted that sample to court last December, but it had allegedly included so much redaction that it made it “unusable,” in the court’s words. The plaintiffs also claimed that OpenAI removed billions of results from ChatGPT after they filed suit in direct violation of the court’s preservation order, and that the AI ​​giant substituted millions of records in the requested sample.

In other words, they claim that OpenAI made it unnecessarily difficult to obtain information that the company had already collected.

“If OpenAI truly believed that copying our clients’ journalism was fair and legal, it would not have hidden the truth about doing so,” Ian B. Crosby, lead attorney for the plaintiffs, said in a statement.

Now, the NYT and The Daily News are asking the judge to sanction OpenAI for allegedly withholding evidence and disrupting the discovery process. They ask the court to stop OpenAI from using the sample of 20 million chat logs as evidence, claiming it is unreliable; accept as a fact that ChatGPT logs would have shown significant regurgitation and substantiation of the plaintiffs’ content; to prevent OpenAI from arguing that the chat logs provided do not demonstrate substantial regurgitation; and make OpenAI pay legal fees for having to pursue this evidence.

In a statement, OpenAI spokesperson Drew Pusateri denied the allegations and accused the Times of trying to access private user conversations as its case weakens.

“As the Times’ case weakens and they are forced to drop their lawsuits against us, they persist in their efforts to invade the privacy of people who have nothing to do with this case, even making these blatantly false accusations,” Pusateri said. “We will continue to uphold our users’ privacy and long-established fair use principles.”