Anthropic’s legal troubles show no signs of abating. Last month, Judge Eumi K. Lee of the U.S. District Court for the Northern District of California denied Anthropic PBC’s second motion to dismiss in Concord Music Group Inc. v. Anthropic PBC, No. 5:24-cv-03811, allowing an octet of music publishers to proceed with copyright and Digital Millennium Copyright Act (DMCA) claims against the AI giant. The lawsuit centers on allegations that Claude, Anthropic’s flagship large-language model, reproduces copyrighted song lyrics without authorization and removes or alters copyright management information (CMI) from its training or output data.
After the original complaint was dismissed, the publishers filed an amended pleading renewing their copyright and DMCA claims. Judge Lee denied a subsequent motion to dismiss the amended claims by Anthropic, accepting as true the allegation that Anthropic knew or should have known that Claude was reproducing copyrighted lyrics since it was trained on large third-party datasets such as Common Crawl and The Pile, which scrape text from across the internet. Notably, the Court touted Anthropic’s own “guardrails” where were intended to prevent infringing outputs as potential evidence that the company may have been aware of the problem and elected to proceed anyway. The Court also found that the Publishers’ claim that Anthropic intentionally trained Claude on an algorithm known as “Newspaper” to remove CMI, such as author names and copyright notices, from those datasets notwithstanding that alternative tools like “jusText” would have retained more CMI, satisfied the threshold for the Publishers’ DMCA claim.
The Court’s decision highlights the growing judicial trend of allowing infringement based challenges to generative AI to proceed past the motion-to-dismiss stage — a trend also reflected in the Bartz v. Anthropic PBC case our IP team wrote on last month. As Concord, Bartz and similar cases move forward, it will be interesting to see whether discovery yields evidence of knowledge, benefit, or intent sufficient to impose liability, and, if so, how courts choose to resolve these cases short of a settlement.
Tannenbaum Helpern’s IP team continues to track the latest developments in major AI and IP litigation.
