OpenAI Copyright Suit Consolidation Portends Consistency, Risk

OpenAI Inc.'s tactical win consolidating a dozen copyright suits against it nevertheless carries risks for the company, as the matters proceed before a judge who’s already ruled against the company in key decisions.

The US Judicial Panel on Multidistrict Litigation last week centralized cases across the country in the US District Court for the Southern District of New York for pretrial activity, which could include dispositive motions including summary judgment, as well as contentious discovery disputes that have been common among the cases.

“This will help create more consistency in the pre-trial outcomes, but it also means that you’ll get fewer tries from different plaintiffs to find a winning set of arguments,” Peter Henderson, an assistant professor at Princeton University, said in an email.

Judge Sidney H. Stein and Magistrate Judge Ona T. Wang, who will oversee the MDL, had already been assigned to six of the lawsuits brought by New York Times Co., Daily News LP, Authors Guild, journalists, and authors. That now includes suits from authors such as Sarah Silverman and Paul Tremblay.

While streamlined, the pretrial proceedings figure to remain contentious as the parties press novel questions about how copyright laws apply to the game-changing generative AI technology. The disputes carry vast ramifications for companies reliant on millions of copyrighted works to train their models.

OpenAI has yet to brief courts on a fair use defense as Meta Platforms Inc. and Anthropic PBC have in other generative AI cases, though it’s said its models are grounded in the principle. Authors and publishers argue there’s no fair use exemption for harvesting protected work to train lucrative commercial technology for free. One court has rejected a fair-use defense for a non-generative AI search tool and fast-tracked an appeal.

Stein’s prior rulings give some insight into his thinking on the issues and where he’s agreed with or diverged from other judges.

For example, he denied OpenAI’s motion to dismiss one Digital Millennium Copyright Act claim brought by Center for Investigative Reporting Inc. and Daily News, while dismissing other DMCA claims from those outlets and The New York Times. He allowed infringement allegations from all three to advance.

Stein’s decision cites one by his district colleague Judge Jed S. Rakoff allowing Intercept Media Inc.'s DMCA claim against OpenAI to proceed. But that clashes with Judge Colleen McMahon’s November order dismissing Raw Story Media Inc.'s claim and her subsequent denial of a bid to amend its allegations. Last week, she told the plaintiffs if they thought Rakoff was right and she was wrong, they should appeal to the Second Circuit.

“Tech companies will want Judge McMahon’s perspective to be the winning view,” Henderson said. “The MDL assignment to Judge Stein throws a wrench in that.”

OpenAI declined to comment, and counsel for plaintiffs in suits transferred to Stein didn’t immediately respond to requests for comment.

Chosen Venue

OpenAI in December requested the cases be transferred to the Northern District of California, arguing it’s where the “overwhelming majority of relevant documents” are located and where the design and training of the large language models took place. Four of the consolidated cases were filed in California, with eight in New York.

Though OpenAI failed to land in its preferred venue, the Second Circuit still has precedent favorable to it, said Santa Clara University law professor Tyler T. Ochoa. He pointed to its 2015 decision in Authors Guild v. Google Inc., which held Google’s copying of books to enable the public to digitally search the texts qualified as fair use under the Copyright Act.

He contrasted that with another Second Circuit case, Andy Warhol Found. v. Goldsmith, that was more favorable to copyright owners.

“The Warhol case specifically distinguishes Google Books, and I think Google Books is much closer factually” to the OpenAI cases, he said.

The New York court may be preferable to plaintiffs, though, because the Northern District of California “is generally viewed as a biz-friendly venue for technology disputes,” said McKool Smith attorney Avery Williams.

Stanford Law professor Mark Lemley, who previously represented Meta in an AI copyright case, agreed, saying in an email that shifting “the center of gravity in the overall cases” to New York creates “significant risk” for OpenAI.

He noted that summary judgment decisions in cases against Meta and Anthropic in California could still be made first.

In any event, the high court will likely have the final word.

“A decision on the fair use standard as applied to generative AI training is destined for Supreme Court review one way or another,” Williams said, adding the OpenAI consolidation “might speed up the process” by teeing it up in one big case.

Plaintiffs’ Challenges

While the the choice of venue may be a silver lining for the plaintiffs, attorneys said there are likely to be coordination issues.

“I do see a potential challenge in amalgamating what seem to be very different kinds of creative content among the plaintiffs,” Bhamati Viswanathan, a professor at New England Law, said in an email. “The news articles of NYT have to clear the fact/expression divide, while the books of Franzen, Gresham, Martin, et al. are closer to the sheerly ‘creative’ that generally gets more scrupulous protection under fair use.”

Attorneys for the copyright owners who filed their lawsuits in California may also have to give up some control to the lawyers for the New York plaintiffs.

“If anybody’s lost here, I think it’s the plaintiffs’ lawyers in the Tremblay and Silverman cases,” Ochoa said. He said he’d expect “plaintiffs in the New York cases will have more of the say about how discovery is conducted.”

Consolidation could increase the likelihood of settlement, though, said Santa Clara Law dean Michael Kaufman.

“All the parties that have common claims are in one forum, in one jurisdiction, in front of one judge, one magistrate,” Kaufman said. “So it’s much more likely to have a global or omnibus settlement.”

The case is In re: OpenAI, Inc., Copyright Infringement Litigation, J.P.M.L., No. 3143.