• by Ray Schultz , Columnist, Yesterday

The New York Times is resisting an effort by OpenAI to get hold of Chat Explorer logs from Times employees for use in discovery in the copyright case filed by the Times, and to set aside an order signed last September by Magistrate Judge Ona T. Wang, saying the logs were not pertinent. 

OpenAI is seeking a reversal of that order and to obtain the logs, according to an argument opposing OpenAI's request on Tuesday.

Chat Explorer logs are an OpenAI product that the Times admits using, but the paper argues that they are irrelevant to this case. 

“OpenAI again attempts to frame this lawsuit as being about The Times’s use of generative AI, not its own massive copyright infringement,” the Times states. “But as this Court has already held, the relevant 'use' in this litigation is OpenAI’s copying of Times content, which includes: (1) OpenAI’s copying of The Times’s works to create internal datasets; (2) OpenAI’s copying of The Times’s works to train GPT models; (3) OpenAI’s copying of The Times’s works to ground results for GPT consumers; and (4) OpenAI’s reproduction, abridgment, hallucination, and misattribution of The Times’s works in outputs that its GPT products display to those third-party consumers.”

That language shows just how nasty this case is getting. Judge Wang has tried to speed up discovery and avoid unnecessary arguments. 

And the explorer logs? 

The documents says that Chat Explorer, which is the Times’s internal interface of the OpenAI API, became generally available to Times employees on December 14, 2023. It is an optional tool for experimentation by Times employees that they were not required to use. The Times eventually stopped adding users to ChatExplorer. 

The Times has turned over certain logs, but nowhere near the hundreds used by 512 employees.

Judge Wang ruled in September that the “relevant ‘uses’ at issue in this case are: (1) Defendants’ alleged copying of [The Times’s] and other plaintiffs’ copyrighted works to (a) create an internal dataset and (b) train GPT models; and (2) the Defendants’ alleged copying of Plaintiffs’ copyrighted material in the outputs of Defendants’ LLMs.”

The issue is not resolved, and OpenAI is certain to come back a robust answer. 

If Judge Wang ever decides to leave the bench, she will be fully qualified to run an AI company.  

The case is on file with the U.S. District Court for the Southern District of New York.