In 2023, several music publishers, including Universal and Concord, filed a lawsuit against Anthropic. The AI company is accused of infringing copyright. Anthropic’s models — on which the bot Claude is based — were trained by the lyrics of songs managed by these publishers. An example is given in the complaint: Claude reproduced almost word for word the lyrics of a Katy Perry song.
Claude in the dock
The problem here is that Anthropic appropriates song lyrics without consent or compensation, while song lyrics aggregators (specialized websites, for example) pay licenses to use these works.
The legal proceedings are ongoing, but an initial agreement between all parties has been validated by the judge in charge of the case. Anthropic agrees to maintain the safeguards that (normally) prevent Claude from producing lyrics for songs owned by publishers, or from generating lyrics inspired by those songs.
The agreement also provides a mechanism for publishers to notify Anthropic if there is an infringement. The company says Claude was not designed to infringe on copyright and reiterates that "we have many processes in place to prevent such infringement." Before adding another layer on the "fair use" that generative AI models should benefit from, the main line of defense of AI companies that train their models on copyrighted content.
The court will make a decision in the coming months on a preliminary injunction that, if issued, will formally prohibit Anthropic from training its future models with copyrighted song lyrics.
Source: Hollywood Reporter
0 Comments