The Up-to-date York Times and The Daily News claim that OpenAI lied about its ability to search customers’ call logs and training datasets for their copyrighted works. It’s the latest escalation in a two-year lawsuit against the artificial intelligence company for allegedly violating copyright law by training its generative artificial intelligence models on Times content and reproducing that journalism in user results.
Throughout the case, OpenAI maintained that it had no way of searching its own training corpus. She also argued that searching or creating ChatGPT’s immense collection of conversations would be technically burdensome and raise user privacy concerns because the logs would need to be downloaded, processed and de-identified. Outlets sought this data to determine whether their copyrighted journalism was present in the OpenAI training dataset and whether and how often ChatGPT generated responses by using or reproducing their content.
In court-mandated April testimony, OpenAI privacy engineer Vinnie Monaco allegedly disclosed that OpenAI had already conducted internal searches and assessments of its training corps for copyrighted journalistic works.
Monaco’s testimony also allegedly revealed that, even before the NYT filed its lawsuit, OpenAI had already amassed a database of approximately 78 million de-identified ChatGPT conversations, which it used internally to determine the extent to which it was infringing the works of others. In addition to this dataset, OpenAI also allegedly implemented a “Bloom” filter as part of a toolkit called “Project Giraffe” that detected and recorded regurgitation in the results shortly after the lawsuit was filed.
These last two revelations are particularly significant. Plaintiffs originally asked OpenAI to provide a sample of 120 million chat logs, but OpenAI negotiated to reduce the sample to just 20 million. OpenAI eventually submitted this sample to the court last December, but it allegedly contained so many redactions that, in the court’s words, the sample became “useless.” The plaintiffs also alleged that OpenAI deleted billions of ChatGPT results after filing the lawsuit in direct violation of the court’s injunction, and that the AI giant substituted millions of logs in the requested sample.
In other words, they argue that OpenAI made it unnecessarily hard to obtain information the company had already collected.
“If OpenAI truly believed that copying our clients’ journalism was fair and legal, it would not conceal the truth that it was doing so,” Ian B. Crosby, lead counsel for the plaintiffs, said in a statement.
Now the NYT and The Daily News are asking a judge to punish OpenAI for allegedly withholding evidence and interfering with the discovery process. They are asking the court to prevent OpenAI from using its 20 million chat log sample as evidence, saying it is unreliable; accept as a fact that the ChatGPT logs would show a earnest lack of closure and grounding of the plaintiffs’ content; to prevent OpenAI from arguing that the provided chat logs do not show significant regurgitation; and making OpenAI pay legal fees for having to pursue this evidence.
In a statement, OpenAI spokesman Drew Pusateri denied the allegations, accusing the Times of trying to gain access to users’ private conversations as the case faded away.
“As the Los Angeles Times’ case weakens and it is forced to drop its claims against us, it continues its efforts to violate the privacy of people who have nothing to do with this case, including by making blatantly false allegations,” Pusateri said. “We will continue to defend the privacy of our users and the long-established principles of fair use.”