AI-Training Digital Copyright Ruling Paves Way for More Lawsuits

The Intercept’s victory over OpenAI Inc.‘s bid to quickly dismiss the publication’s Digital Millennium Copyright Act lawsuit marks the first such case to clear initial court hurdles, setting the stage for similar claims against AI companies.

As lawsuits pile up over the processes used to train AI models, plaintiffs are getting smarter about how to bring claims against the novel technology, attorneys said. After other DMCA-focused complaints against AI firms were dismissed, Judge Jed S. Rakoff’s Nov. 21 ruling provides insight into what courts need to see to move those claims forward, said Bhamati Viswanathan, assistant professor at New England Law.

“Judge Rakoff is signaling that this door isn’t shut” on DMCA claims, she said. “That’s really interesting, because right now,” she added, “we’re at the ‘throw everything at the wall and see what sticks’ point.”

Intercept Media Inc. included in its amended complaint allegations that specific sets of algorithms OpenAI has admitted to using—including one called Dragnet—can pull the entire copy of an article and strip it of “copyright management information” like the author’s name and story title. That, The Intercept alleged, is a violation of the DMCA, which bars intentionally removing CMI.

Rakoff has yet to issue an opinion explaining his reasoning. If his rationale for letting the DMCA claim proceed is based on that evidence, it sets up a clear pathway that can be used by other plaintiffs who can point to the same information, said Peter Henderson, assistant professor at Princeton University.

“Plaintiffs seem to be doing a better job of investigating down to the tools that were used to download and process the data,” he said. “This may be a formula, going forward, for getting to discovery.”

DMCA Pathway

DMCA claims haven’t fared as well in other cases.

Earlier this year, in Doe 1 v. Github, Judge Jon S. Tigar dismissed programmers’ claims, saying they failed to show OpenAI had reproduced their code identically. And earlier this month, another New York federal judge, Colleen McMahon, nixed claims from online publications Raw Story and Alternet that OpenAI violated the DMCA by stripping CMI from their articles, saying the they hadn’t alleged concrete harm.

An opinion is pending in The Intercept’s case, but Rakoff’s order in the US District Court for the Southern District of New York made clear the type of claim he felt should proceed: that OpenAI intentionally removed CMI knowing it would lead to copyright infringement. The Intercept’s other claims—that OpenAI distributed copyrighted content stripped of CMI, and that Microsoft Corp. is guilty of the same activities—were dismissed.

The ruling, however, isn’t a “sweeping rejection of the harm theory espoused in” the Raw Story case, said Aaron Moss from Greenberg Glusker Fields Claman & Machtinger LLP.

“As Judge Rakoff indicated in the hearing on this motion, this really seems to be an issue of fact, not an issue of law,” Moss said. “Based on the allegations, the conservative route for a judge to take in a case like this is to say, ‘Let’s go to discovery.’ That’s not to say that at the end of the day it will prevail.”

Though the suit will proceed, The Intercept will now need to prove that OpenAI removed CMI either knowing—or having reasonable grounds to know—that ChatGPT would essentially regurgitate The Intercept’s articles, he said.

Ultimately, the case will come down to what is revealed in the discovery process.

“So far, we have been limited to basing our claims on publicly available information, and we’re confident that there is a significant amount of additional information that we’ll be able to uncover and discovery about exactly how OpenAI went about doing this and their knowledge that it would conceal or facilitated copyright infringement,” said Loevy & Loevy’s Matt Topic, attorney for The Intercept and Raw Story Media.

In an interview last week, Topic called the DMCA the “only real legal recourse” for digital publishers without copyright registrations for each article—like The Intercept—to push back on the unlicensed use of their materials. Some online publications don’t have registrations because registering individual articles was an onerous process, though the Copyright Office issued a final rule this year allowing group registrations for news websites.

The DMCA pathway, while a boon for online outlets, is also a potential headache for OpenAI, Henderson said.

“OpenAI may want to appeal one of these decisions to try to make sure that these 1202(b) claims don’t keep going forward to discovery,” Henderson said, referring to the CMI provision of the DMCA.

OpenAI’s attorneys didn’t respond to a request for comment.

Transparency Effects

Rakoff’s order has few direct implications for the bulk of copyright lawsuits against AI companies that, unlike The Intercept’s, allege direct copyright infringement.

But as rulings begin to trickle in, their impact could broaden—and possibly make discovering what’s being used for training data in newer AI models more difficult.

“The side effect of this might be that model creators further lock down any information about how they downloaded the data, how they process the data,” Henderson said. “Previously it was like, ‘Well, we’ll just hide the data itself.’ But now model creators might be incentivized to not reveal any pre-processing they did to that data.”

Regulators have launched efforts to bolster transparency in the AI industry, with California governor Gavin Newsom signing a bill last month to require AI firms to disclose how their models were trained. The European Union’s Artificial Intelligence Act also includes a provision that requires AI companies to make public a “sufficiently detailed summary” about their training data starting August 2025.

Such measures to curtail secrecy around training data could prove important.

AI firms concealing what data they used to train their models would “not be ideal from a reproducibility or transparency standpoint,” Henderson said. “And it certainly doesn’t solve creators’ concerns.”

The case is The Intercept Media, Inc. v. OpenAI, Inc., S.D.N.Y., No. 1:24-cv-01515.