Newspaper publishers hash out discovery with OpenAI, Microsoft in copyright suit

MANHATTAN (CN) — OpenAI lost a bid to make The New York Times reproduce ChatGPT results that the outlet says amount to copyright infringement on Thursday as the artificial intelligence company met with newspaper publishers in Manhattan federal court to iron out document production issues.

The New York Times sued OpenAI and Microsoft last December claiming the AI developers drew from large swaths of copyrighted articles to train large language models that enhance the ability of ChatGPT and Copilot to generate text in a variety of styles.

Regional newspapers owned by MediaNews Group and Tribune Publishing companies filed similar claims in April. Those outlets include Tribune Publishing’s Chicago Tribune, Orlando Sentinel and South Florida Sun Sentinel and the New York Daily News, as well as Media News Group-owned Mercury News, Denver Post, Orange County Register and St. Paul Pioneer-Press.

The Authors Guild and Center for Investigative Reporting, which have filed separate lawsuits under similar claims, were present for Thursday’s four-hour conference to address discovery issues in all four cases — but the proceedings focused largely on disputes between The New York Times and the AI developers.

OpenAI asked the court to compel The New York Times to turn over documents showing its “failed attempts” at producing the outlet’s articles verbatim when prompted by ChatGPT. According to The New York Times’ complaint, that outcome requires “minimal prompting.”

“We understand that your theory is that people are going to go away from the Times in droves,” Andrew Gass, an attorney with Latham & Watkins representing OpenAI, said Thursday. To counter that claim, OpenAI wants the Times to demonstrate how many tries it took to generate the examples listed in the outlet’s complaint.

Katherine Peaslee, an attorney with Susman Godfrey representing The New York Times, said those efforts aren’t relevant to the case now. “The Times is not going to seek to rely on it,” Peaslee said. “It’s about how the average user uses ChatGPT.”

She argued that the AI developers’ use of its content to train its large language models still constituted copyright infringement. “Negative outputs don’t disprove that,” Peaslee said.

U.S. Magistrate Judge Ona T. Wang denied OpenAI’s motion from the bench.

OpenAI also requested that The New York Times produce documentation to prove that the works used in ChatGPT are “original works of authorship.”

Judge Wang took issue with the suggestion that journalists should have to quantify how much of their work is original and how much comes from outside sources, like quotes or citations.

“The goal is to understand what portion of the copyrighted works The Times owns,” Elana Dawson, an attorney with Latham & Watkins also representing OpenAI, said Thursday. “We have to be able to determine whether or not those works are owned by The Times.”

Davida Brook, an attorney with Susman Godfrey also representing The New York Times, responded that OpenAI’s request is “not appropriate,” and countered that the publisher was willing to produce its copyright registration for certified works and produce proof of contractual relationship with authors of the works in question.

“This has been an enormous endeavor because we are talking about 10 million plus works,” Brook said.

While Wang also denied that motion, she allowed OpenAI to renew it if it tells The New York Times which works specifically were used to train its large language models.After the hearing, Brook praised the judge’s decision. “Today, the court denied OpenAI’s demand that The Times turn over all of its reporters’ files for the millions of articles that OpenAI copied, without permission or payment, to build and operate their commercial tools.” Brook said in a statement. “This is a win for press freedom and for all copyright owners who seek to protect the work that they create.”

For the remaining discovery issues, Wang asked the parties to confer and come back with an update by Sept. 20.

Follow @NikaSchoonover