Round Hill sues Suno and Anthropic over alleged music scraping as leaked code reveals training data

Independent music publisher Round Hill Music has filed separate copyright lawsuits against Suno and Anthropic, seeking damages exceeding $1 billion. The complaints allege unauthorized scraping of copyrighted compositions, recordings, and lyrics to train AI models. This litigation comes amid a major data leak exposing Suno's scraping of over 2 million YouTube songs via proxy firm Bright Data. While Suno and Anthropic assert fair use defenses, Round Hill's legal team is challenging the permanent stockpiling of copyrighted works, building on a complex landscape of industry licensing deals and ongoing union disputes over artist compensation.

Round Hill copyright lawsuits

• ▪The Round Hill Music lawsuit against Suno covers both compositions and recordings, seeking damages that could exceed $1 billion.
• ▪Round Hill Music filed two separate copyright lawsuits on August 17, 2026, alleging that Suno and Anthropic trained AI models on its copyrighted music without a license.
• ▪Round Hill Music's initial lawsuits name roughly 500 songs as a bellwether sample, with plans to expand the cases to cover 10,000 or more compositions.
• ▪The Round Hill Music lawsuit against Anthropic covers only lyrics and compositions, alleging that Anthropic's Claude text model was trained on copyrighted lyrics without a license.

Unauthorized music scraping methods

• ▪Suno routed its scraping through proxy firm Bright Data to bypass YouTube's defenses and hunt for clean, a cappella versions of songs.
• ▪Round Hill Music's lawsuit against Suno names Israeli data-scraping company Bright Data as a defendant, alleging it supplied proxy networks and scraping tools to pull music off licensed platforms.
• ▪Round Hill Music's lawsuits allege that Suno and Anthropic stripped out embedded copyright management information identifying the owners and terms of use during scraping.

Leaked Suno training data

• ▪A hacker using the alias ellie.191 claimed to have breached Suno using the Shai-Hulud worm to steal employee credentials, customer emails, phone numbers, and Stripe records.
• ▪Leaked source code obtained from a hacker who breached Suno reveals that the AI music tool trained its model by scraping millions of songs and lyrics across the internet.
• ▪Suno characterized the November 2025 data breach as limited and quickly contained, stating that the exposed code was outdated and no sensitive customer data was leaked.
• ▪The leaked Suno data included a file labeled 'youtube_music' containing more than 2 million clips, alongside tens of thousands of hours pulled from Deezer, Genius, and Pond5.

Fair use defense arguments

• ▪Round Hill Music's lawyers argue that maintaining a permanent stockpile of copyrighted material for future AI training fails every factor of the American fair-use test.
• ▪A Munich court ruled against Suno on July 31, 2026, in a case brought by German collecting society GEMA, rejecting Suno's defense that training did not require a license.
• ▪Suno defends its training practices as fair use, arguing that it trains on publicly available music files to build models intended for original creation.

Industry licensing deals

• ▪Independent music publisher BMG chose to bypass litigation entirely, establishing a direct licensing alliance with Suno.
• ▪Warner Music settled its copyright litigation with Suno in November 2025, exiting the lawsuit in exchange for a licensing agreement.
• ▪The American Federation of Musicians sued Universal Music and Warner Records on June 5, 2026, alleging the labels failed to pay session musicians from their AI licensing settlements.