Anthropic faces lawsuit over use of Elvis Presley music in AI training

The lawsuit against Anthropic

Anthropic is facing a new legal challenge from the owners of Elvis Presley's music. The lawsuit, filed on Friday, alleges that the company used the singer's catalog to train its artificial intelligence models without obtaining the necessary permissions.

The plaintiffs include Authentic Brands Group, which owns the rights to Presley's music, along with several other entities that hold rights to the "Jailhouse Rock" singer. These groups are seeking to enforce their exclusive rights against the AI developer.

As reported by The Hollywood Reporter, the legal action follows a growing trend of music industry giants taking action against AI firms. The lawsuit specifically targets the way these models are built and the scale at which they ingest copyrighted data.

Specific songs and claims

The complaint lists 274 specific songs from the Elvis Presley catalog that were allegedly infringed. Notable tracks included in the filing are "Can't Help Falling in Love," "A Little Less Conversation," and "Don't Be Cruel."

The plaintiffs argue that the AI model, known as Claude, has created infringing output for nearly every song available on the internet. They claim the technology is capable of reproducing the core elements of these copyrighted works rather than just being inspired by them.

Lead attorney Richard Busch stated in the filing that the plaintiffs are "deeply protective of these iconic songs." He further argued that the company has built its entire business model on copyright infringement at a scale that the industry has never seen before.

Technical output and guardrails

The lawsuit highlights specific instances where the AI model appeared to bypass safety measures. According to the filing, even after Anthropic supposedly implemented guardrails, the model continued to produce infringing content when prompted to recreate songs for a modern audience.

The plaintiffs point to a specific interaction where the model admitted to keeping the same structure, hooks, and full sets of verses and bridges. The suit claims the AI was merely swapping out words while keeping the underlying copyrighted composition intact.

This suggests that the model's ability to mimic the specific creative choices of a songwriter remains a primary concern for rights holders. The legal team argues that these outputs edge past simple inspiration and move into the territory of reproducing copyrighted material.

Financial stakes and history

The legal action carries significant financial weight, as the plaintiffs are seeking $150,000 for every song they claim was infringed. This would represent a massive payout if the court finds the AI company liable for the entire list of 274 tracks.

This is not the first time Anthropic has faced these types of accusations. The company was previously sued by Universal Music Publishing, Abcko, Concord, and BMG for similar claims regarding the use of lyrics to train its AI agent.

In August, Round Hill also filed a lawsuit against both Anthropic and Suno. That suit alleged that the AI companies used hundreds of songs from a publisher's catalog to train their models without permission, showing a broader industry pushback.

What comes next for AI

Anthropic has not yet provided a public comment regarding the new lawsuit. The company's response will likely determine how it plans to defend its training methods and whether it will seek to license music in the future.

The outcome of this case could influence how other AI developers handle copyrighted data. If the court rules in favor of the music owners, it may force companies to pay significant licensing fees or fundamentally change how they train their models.

For now, the music industry continues to monitor these cases closely. The goal for the plaintiffs is to establish a clear boundary between using data for learning and using data to create competing commercial products.