Judge to hear Anthropic copyright defense in lyrics lawsuit

Copyright disputes involving new reproduction technologies have historically followed a pattern where courts must determine if technical copying serves a different social function than the original work. Precedents such as Authors Guild v. Google Books (2015) established that digitizing books for search indexing could be transformative fair use even when entire works were copied, provided the output did not serve as a market substitute. Similarly, Sega v. Accolade (1992) found that reverse-engineering software code for interoperability purposes was fair use despite literal copying during the process. These cases turned on whether the defendant's use added new utility or merely repackaged existing expression. In each instance, courts balanced the four statutory fair use factors rather than applying bright-line rules about copying volume. This judicial approach creates uncertainty for emerging technologies where the line between functional learning and expressive reproduction is technically complex and fact-specific, requiring case-by-case adjudication rather than categorical exemptions.

On October 22, 2026, US District Judge Eumi Lee presided over a pivotal hearing in San Jose, California, regarding a copyright infringement lawsuit filed against AI developer Anthropic by Universal Music Group (UMG) and other music publishers. The proceeding focused exclusively on the legal sufficiency of Anthropic’s affirmative defense of fair use concerning the ingestion of copyrighted song lyrics to train its Claude chatbot model. According to Reuters Legal, Judge Lee did not issue an immediate ruling from the bench but instead engaged both parties in rigorous questioning regarding the application of fair use doctrine to generative AI training methodologies.

The plaintiffs, led by UMG, allege that Anthropic unlawfully exploited millions of copyrighted song lyrics without obtaining licenses or providing compensation to rights holders. According to filings referenced in Reuters reporting, the publishers argue that this ingestion constitutes direct commercial exploitation rather than transformative research, asserting that Anthropic’s products compete directly with the creative works they were trained upon. The lawsuit claims that allowing such unlicensed use would undermine the economic incentives for human creators and set a damaging precedent for intellectual property rights in the digital age.

In response, Anthropic has urged the court to rule that its training practices are protected under US copyright law as fair use. According to Reuters Legal, Anthropic argues that the processing of lyrics is fundamentally transformative, serving the functional purpose of teaching language patterns and semantic relationships rather than reproducing artistic expression for consumption. The defense contends that the model’s output does not substitute for the original songs and that restricting access to training data would stifle innovation in artificial intelligence development. Anthropic seeks to have the lawsuit dismissed based on this interpretation of fair use protections.

During the October 22 hearing, Judge Lee pressed attorneys for both sides on specific aspects of the fair use analysis. According to Reuters and Straits Times reporting, the judge’s inquiries suggested a careful examination of whether AI training qualifies as sufficiently distinct from traditional copyright infringement precedents. The court’s line of questioning indicated that the resolution may hinge on technical distinctions between memorization of protected expression and the extraction of non-protectable statistical patterns. No timeline was provided for when a written opinion might be issued, leaving the industry awaiting clarification on how federal courts will apply decades-old copyright frameworks to novel machine learning technologies.

This hearing represents a critical juncture in ongoing litigation between content creators and AI developers. While previous cases have addressed issues like web scraping and database copying, the specific question of whether large-scale model training constitutes fair use remains largely unsettled in federal jurisprudence. According to multiple news outlets covering the event, the outcome of this case could influence dozens of similar lawsuits currently pending against major model developers across various jurisdictions. Both UMG and Anthropic presented their arguments through established legal counsel, with the proceedings adhering to standard civil procedure despite the novel technological subject matter at issue.

Universal Music Group and co-plaintiffs seek to establish that AI training requires licensing, potentially creating new revenue streams from millions of allegedly infringed lyrics. Anthropic faces existential risk if fair use defense fails, as training data restrictions could impair Claude's language capabilities and expose the company to substantial damages. Dozens of parallel lawsuits against major model developers await this ruling's reasoning, meaning the decision could reshape AI development economics industry-wide. Content creators stand to gain leverage in licensing negotiations if courts reject broad fair use interpretations for training data ingestion.