San Francisco's Anthropic, the multibillion-dollar AI company known for its Claude chatbot, is facing a federal lawsuit from a group of music publishers who say the company used huge amounts of copyrighted song lyrics to train its product.

Chatbots like Claude and OpenAI's ChatGPT interface are fed an enormous number of language examples from across the internet, and their programming allows them to base answers to user prompts off of what they have "read."

Anthropic did not immediately respond to an emailed request for comment.

The company has billed itself as building "Constitutional AI" tools, saying its chatbots are less likely to make things up or otherwise go off the rails because of a central set of programming rules that they abide by.

Last month Amazon said it would invest up to $4 billion in the company.

The suit was filed in federal court in Tennessee by Universal Music and more than a dozen other music publishers. The plaintiffs are seeking damages and a court order requiring Anthropic to stop the practice.

The arguments by the plaintiffs are similar to those made by attorneys for authors including Michael Chabon and David Henry Hwang in previous, separate lawsuits against OpenAI and Facebook parent Meta, essentially, that you can ask the chatbots for information about a title or song and it can answer, so the work must have been part of the training data.

The music publishers' attorneys said in a filing that when Anthropic's Claude chatbot is asked to provide the lyrics to songs like "A Change Is Gonna Come," "God Only Knows," "What a Wonderful World," "Gimme Shelter," "American Pie" and many others, it will provide responses that contain all or significant portions of those lyrics.

Because Anthropic did not pay to license that material for commercial use, the company violated the law, the complaint argued.

Some experts have raised doubts about whether using copyrighted material as part of a training set violates copyright law, saying that the protected works are not part of the actual programming of the model itself, and that instead it merely learned from them.

Lawyers for comedian Sarah Silverman also brought a similar suit against Meta's LLaMA AI model.

Attorneys for the social media giant in that case pointed to another in which Google Books was found not to have violated copyright law when scanning and indexing countless works to make them searchable, also saying the company's chatbot code does not contain any copyrighted works itself.

Reach Chase DiFeliciantonio: chase.difeliciantonio@sfchronicle.com; Twitter: @ChaseDiFelice