Elvis Presley Estate Sues Anthropic for $48.6 Million Over AI Song Training

Artificial intelligence developers routinely argue that their large language models learn from copyrighted texts the way a human student does, absorbing patterns rather than memorizing exact words. But a new lawsuit from the estate of Elvis Presley sets that defense against the words of Anthropic’s own chatbot, Claude, which reportedly admitted its song rewrites “keep landing very close to the originals.”[2][4]

The complaint, filed on October 23, 2026, in the US District Court for the Northern District of California, accuses Anthropic of training its models on 274 of Presley’s songs without permission. The plaintiffs, led by a unit of Authentic Brands Group, are seeking roughly $48.6 million in total damages.[4][7]

Lead attorney Richard Busch of Adams and Reese wrote in the filing that Anthropic has built a business model on copyright infringement “at a scale never seen before.” The suit names iconic tracks including “Can’t Help Falling in Love,” “Don’t Be Cruel,” and “A Little Less Conversation.”[2][8]

The Presley estate’s legal action represents a significant escalation in the ongoing tension between the music industry and artificial intelligence firms. While major record labels have previously sued AI developers, this marks one of the most prominent artist estates directly challenging a leading foundation model.[1][7]

The Output Problem

The lawsuit does not merely focus on the ingestion of lyrics during the model’s training phase, but heavily targets the specific outputs Claude generates. The plaintiffs acknowledge that Anthropic implemented guardrails to prevent the chatbot from printing verbatim song lyrics when directly prompted.[2][4]

However, the complaint alleges those safeguards fail when users ask Claude to rewrite the classic songs for a modern audience. According to the filing, the AI model returns versions that maintain the original structure and hooks, swapping out just enough words to bypass the filters.[2][4]

The plaintiffs included a striking piece of evidence: Claude’s own self-assessment of its generated lyrics. The lawsuit quotes the chatbot acknowledging that its results go beyond mere inspiration, stating that the renders feature the “same structure, same hooks reused with light rewording, and full sets of verses/choruses/bridges.”[4]

This distinction between training data and generated output is central to the legal strategy. By demonstrating that Claude can essentially reconstruct the copyrighted works on demand, the estate aims to undermine the argument that the model only absorbed abstract stylistic concepts.[3][7]

Calculating the Damages

The financial penalties sought in the case reflect the aggressive stance taken by the rights holders. The plaintiffs are asking for the maximum statutory damages of $150,000 per willfully infringed song, which totals approximately $41.1 million for the 274 tracks listed in the complaint.[4][5]

Additional claims push the total figure higher. The lawsuit seeks $685,000 for circumvention violations, calculated at up to $2,500 per instance, and another $6.85 million for credit-stripping claims at up to $25,000 per violation, bringing the total demand to roughly $48.6 million.[4]

Beyond monetary compensation, Authentic Brands Group is asking the federal court for a permanent injunction against Anthropic. The plaintiffs want an order forcing the company to purge all unlicensed Presley lyrics from Claude’s training data and underlying model weights.[4]

The estate is also demanding a full accounting of Anthropic’s scraping activities and training datasets involving the catalog. This discovery process could force the AI company to reveal exactly how it sourced the lyrics, which the plaintiffs allege came from Common Crawl, The Pile, and pirate libraries.[4]

A Broader Legal Pattern

The Presley lawsuit joins a growing pile of litigation facing Anthropic over its use of copyrighted music. Earlier in the week, Busch filed a similar complaint against the company on behalf of Reservoir Media, following an August 2026 lawsuit from music publisher Round Hill.[7]

Universal Music Publishing, Concord, and BMG have also previously taken legal action against Anthropic, claiming the company used their lyrics to train Claude. These coordinated legal challenges suggest the music industry is systematically targeting foundation models over their text-ingestion practices.[2]

The outcome of these cases will likely hinge on whether federal courts classify the ingestion of copyrighted lyrics for AI training as fair use. If judges rule against the AI developers, companies like Anthropic could face massive retroactive licensing fees or be forced to rebuild their models entirely.[7]

For now, the Presley estate’s lawsuit ensures that one of the most recognizable catalogs in American music history will serve as a test case. The next legal milestone will be Anthropic’s formal response to the October 23 filing, which will outline the company's defense strategy.[6][7]