Elvis catalog owners sue Anthropic for $41 million over AI training

Elvis Presley music stakeholders sued Anthropic for $41 million over alleged training on 274 songs. The single-artist case shows copyright exposure for AI content workflows and licensing.

What happened

The owners of Elvis Presley’s music sued Anthropic for $41 million, The New York Times reported. The lawsuit claims the AI company trained its models on 274 Elvis songs without permission. Stakeholders in the catalog include Authentic Brands Group, Raleigh Music Publishing, multiple trusts and at least one individual.

The complaint, filed on October 23rd, accuses Anthropic of bypassing DRM protections and removing or altering copyright management information from 247 songs. It names hits such as Don’t Be Cruel, Can’t Help Falling in Love and A Little Less Conversation.

"The Elvis Plaintiffs bring this action to enforce these exclusive rights against the Defendant that has built its entire business model based on copyright infringement at a scale never seen before," lead attorney Richard Busch wrote.

The suit is part of a broader pattern. Sony and Warner recently sued Anthropic over thousands of works, and Universal Music and Concord Music Group previously filed a claim involving 20,000 copyrighted songs. Anthropic has said that training on copyrighted music is "fair use" and intends to defend itself.

Key facts

• Stakeholders said Anthropic trained its models on 250-plus Elvis songs and demanded $41 million in compensation.
• The complaint alleges training on 274 Elvis songs without permission and names hits like Don’t Be Cruel, Can’t Help Falling in Love and A Little Less Conversation.
• It says Anthropic bypassed DRM protections and removed or altered copyright management information from 247 songs.
• The suit was filed on October 23rd by Authentic Brands Group, Raleigh Music Publishing, multiple trusts and at least one individual.
• The complaint notes Elvis sold over a billion records and describes his works as "some of the most valuable in the world."
• Prior Anthropic copyright claims involve 20,000 songs from Universal Music and Concord, and thousands of works from Sony and Warner; each lawsuit claims damages in the billions of dollars.

Our analysis

The single-artist framing is the notable part. Elvis died nearly 50 years ago, yet the claim is $41 million. That suggests rights holders see AI training as a per-catalog legal and financial exposure question, not just a dispute over current stars. If a decades-old catalog can support a claim this size, operators who produce or prompt AI music content likely face similar demands across older catalogs.

The DRM and copyright management information allegations matter because they go beyond the "fair use" debate. The lawsuit claims Anthropic removed or altered identifying information from 247 songs. That kind of allegation, if proven, could weaken a fair-use defense and likely makes settlement pressure more direct. For platforms, this is a signal that technical access and metadata choices may be litigated as much as the act of training.

Anthropic’s existing disputes with Sony, Warner, Universal and Concord show that AI model developers are being approached by both publishers and labels. The Elvis suit is smaller in song count than the 20,000-song claim, but its focus on one legacy artist makes it easier for the public and courts to understand. That may increase reputational pressure and likely prompt more conservative licensing behavior across AI tools.

What it means for operators

• Audit AI-generated music or lyric output before publishing. A prompt that returns recognizable Elvis-style or lyric fragments may create copyright risk, especially with older catalogs.
• Do not treat "fair use" as settled. If your AI content tool can output protected lyrics or metadata, document licensing and restrict prompts that target named artists.
• Track rights-holder claims and platform policy changes. The Sony, Warner, Universal and Concord cases suggest publishers and labels are now coordinating around AI training.
• For creators using AI audio tools, favor licensed or original training sources and keep records of model versions and prompts used, since metadata alteration allegations could extend liability beyond the model developer.