Anthropic Awaits Rulings on Lyrics and the Pentagon

SAN JOSE, Calif., Oct. 21 — AA federal judge in San Jose heard several hours of argument on Wednesday in music publishers’ suit over Anthropic’s use of the lyrics of 499 songs, and did not rule. Judge Eumi K. Lee took the motions under submission without signaling her view, according to Courthouse News, the only account of the hearing Daily Cortex found.

In Washington, about a dozen groups and individuals, among them employees of OpenAI and Google in their personal capacities, asked an appeals court this week to let them back Anthropic’s bid to reopen its fight with the Department of War.

Both cases follow Anthropic toward a listing: its public prospectus, not on the S.E.C.’s EDGAR system by early Thursday, could come as soon as Friday, Yahoo Finance reported, citing three people familiar with the plans.

The publishers, among them Concord and ABKCO, say Anthropic copied their works in training Claude and in its output, and that the copying “is not fair use.” They say they already license their lyrics for use with AI, and that Anthropic can pay for a license or stop training on them.

Anthropic calls training on lyrics “transformative” fair use, citing Bartz v. Anthropic and Kadrey v. Meta, two 2025 rulings finding book training fair use. It says Claude’s users, not Anthropic, are directly liable for copied outputs. Neither ruling settles this case: Bartz involved no allegation of infringing outputs. Kadrey rejected lost fees for licensing transformative AI training as market harm, but called market dilution, which the publishers press here, a “potentially winning argument” the authors failed to support with evidence.

Matthew Oppenheim, for the publishers, said songs Claude writes compete with human artists’ work. “It’s not curing cancer, it’s to create other lyrics,” he said, as quoted by Courthouse News. Asked by the judge whether his argument would give publishers a monopoly over lyrics, he said, “Copyright is a monopoly by nature.” Sonal Mehta, for Anthropic, said: “There isn’t anything in the record that shows doom for the music industry.”

The publishers also say Anthropic removed identifying information from their lyrics, in violation of the Digital Millennium Copyright Act. In briefs the judge requested, Anthropic said a Ninth Circuit ruling, Doe 1 v. GitHub, “forecloses” that claim. The publishers answered that GitHub “does not speak to” removal during training and that its analysis “supports Publishers’ claims” on outputs, saying Anthropic’s models had output “verbatim and near-verbatim copies of lyrics to each of Publishers’ 499 Works in Suit.”

On the day of the hearing, Reservoir Media Management, Inc., whose name matches that of Nasdaq-listed Reservoir Media’s operating company, sued Anthropic in the same court; Daily Cortex has not seen the complaint.Six SuitsMusic companies have six copyright suits against Anthropic in the San Jose court: this one, a second by Concord (Jan. 28), BMG (March 17), Round Hill (Aug. 17), Sony Music Publishing (Aug. 28) and Reservoir. Judge Lee has the first five.

In the Washington case, the Secretary of War designated Anthropic a supply-chain risk under a 2018 procurement law, barring it from bidding or subcontracting on certain department technology procurements. The department acted, the appeals court wrote, “after Anthropic refused to relax contractual prohibitions on the use of Claude for lethal autonomous warfare or domestic surveillance.”

On Sept. 25 a divided panel upheld the designation. Judge Gregory G. Katsas, for the majority, found “ample support” for the department’s conclusion: restrictions Anthropic builds into Claude had more than once stopped it from doing tasks government users requested, and a dispute had arisen over whether its contract allowed Claude’s use in an overseas military operation.

Judge Karen LeCraft Henderson, dissenting, wrote that the next contractor the Secretary pressed would face a choice: “Agree to the Secretary’s demands or risk being designated a national security threat.” Anthropic’s Oct. 13 petition says the majority “transformed a statute aimed at hostile foreign actors into one that presumptively reaches every government contractor.”

On Monday the government asked for 14 more days, to Nov. 13 instead of Oct. 30, to respond, citing its lawyers’ other deadlines; Anthropic took no position. The motion restates the department’s concern that Anthropic “might manipulate its model to enforce its own policy judgments about the appropriate ways for the Department to use the model.”

The OpenAI and Google employees wrote that frontier AI used for “domestic mass surveillance” or lethal autonomous weapons “without human oversight” presents risks that “require some kind of guardrails.”

Daily Cortex is written with AI models, including Anthropic’s Claude, and fact-checked with OpenAI’s GPT models.