Anthropic Publishes AI Alignment Research as It Faces New Music Publisher Lawsuit
aiinsider ·
Anthropic made news this week for both a research milestone and a legal conflict. The company published a paper describing an automated system that improves another AI model's alignment without human guidance, beating proposals from experienced human researchers within six hours at a much lower cost. The study acknowledged that results depend on how well benchmarks reflect true alignment goals, and it is viewed as a step toward recursive self-improvement. Separately, Anthropic and two co-founders were sued by Sony Music Publishing, Warner Chappell, and other publishers for allegedly torrenting and scraping copyrighted material to train its Claude models. The lawsuit was filed in the U.S. District Court for the Northern District of California, and Anthropic said it disagreed with the claims and would defend itself. The case follows earlier copyright litigation, including a $1.5 billion settlement in the Bartz v. Anthropic case, where training on copyrighted material was ruled legal but acquiring it through piracy was not. The new suit expands on those claims, citing alleged piracy of millions of books containing song lyrics and sheet music.
aiinsider