The New York Times and Daily News allege that OpenAI lied about its ability to search customers’ chat log data and copyrighted training datasets. This is the latest escalation in a two-year lawsuit against the AI company for allegedly violating copyright law by training generative AI models on Times content and reproducing its journalism with user output.
Throughout the lawsuit, OpenAI has argued that it lacked the ability to search its own training corpus. They also argued that searching or creating large collections of ChatGPT conversations is technically demanding and raises user privacy concerns because logs must be captured, processed, and anonymized. The news organization requested that data to determine whether copyrighted journalism was present in OpenAI’s training dataset and whether and how often ChatGPT used or reproduced that content to generate responses.
Vinnie Monaco, OpenAI’s data privacy engineer, allegedly disclosed in a court-ordered deposition in April that OpenAI had already conducted an internal review and evaluation of its training corpus to search for copyrighted journalistic works.
Monaco’s deposition also revealed that even before the NYT filed its lawsuit, OpenAI had already amassed a database of approximately 78 million anonymized ChatGPT conversations, which it used internally to determine how much it was infringing on the copyrighted works of others. In addition to that dataset, OpenAI allegedly implemented a “Bloom” filter as part of a suite of tools called “Project Giraffe” shortly after the lawsuit was filed to detect and preserve records of backflow in the output.
The last two revelations are especially important. The 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 ultimately submitted its sample to the court in December, but it contained so many edits that it was, in the court’s words, “unusable.” Plaintiffs also alleged that OpenAI deleted billions of ChatGPT outputs after filing the lawsuit in direct violation of court retention orders, and that the AI giant replaced millions of logs with requested samples.
In other words, they claim that OpenAI made it unnecessarily difficult to obtain information that the company was already collecting.
“If OpenAI truly believed that copying its clients’ journalism was fair and legal, it would not have hidden the truth about what it did,” Ian B. Crosby, lead attorney for the plaintiffs, said in a statement.
Now, the NYT and Daily News are asking a judge to discipline OpenAI for allegedly withholding evidence and obstructing the discovery process. They are asking the court to prevent OpenAI from using 20 million chat log samples as evidence as unreliable. Accept as a fact that the ChatGPT logs would have shown significant reflux and basis for Plaintiff’s content. To prevent OpenAI from claiming that the provided chat logs do not show significant backflow. And make OpenAI pay the legal costs of pursuing this evidence.
In a statement, OpenAI spokesperson Drew Pusateri denied the allegations and accused the Times of trying to access users’ private conversations as the allegations faded.
“As the Times’ case weakens and they are forced to drop their claims against us, they continue to persist in their efforts to violate the privacy of people who had nothing to do with this case, including by making blatantly false claims like this one,” Pusateri said. “We remain committed to user privacy and long-established principles of fair use.”
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
