Nashville Musician Finds 71% of Discography Used by AI Models Without Permission

Adam Paddock, a Cincinnati‑born musician who has been recording for nearly a decade, discovered that 71 percent of his catalog is being used by artificial‑intelligence companies to train music‑generation models without his permission. The revelation came after Paddock searched the Atlantic’s newly released AI Watchdog database, which lists the songs that have been harvested for AI training.

Paddock’s discovery was made public on June 16, 2026, when he found 48 of his tracks listed in the database. According to the database, which contains more than 21 million songs used by AI developers, Paddock’s work represents a significant portion of his output. He explained that each song required “100 hours plus invested in it directly, let alone indirectly,” and that the time and effort he puts into creating music is far greater than the time AI companies spend downloading and uploading his recordings to their training pipelines.

The AI Watchdog project, launched by The Atlantic in September 2025, was designed to expose the content used to train large‑scale AI models. The database is searchable and includes four separate collections: two with 12 million and 9 million tracks, and two smaller sets of about 100 000 songs each. The Atlantic’s investigation has highlighted that many independent artists, including Paddock, have had their music incorporated into AI training sets without consent or compensation.

Independent artists across the country report similar experiences. Paddock noted that friends have seen “hundreds of songs stolen” and that “many others with dozens of songs taken.” The lack of a mechanism to prevent or reverse the use of their recordings has left artists without payment. “We’re getting to see no money from that,” Paddock said, emphasizing that producers, writers, instrumentalists, and engineers all lose revenue when their work is used without authorization.

Legal and industry responses are mounting. Class‑action lawsuits and petitions have been filed against AI companies such as Suno and Udio, which are known to offer generative music‑creation services but do not disclose the datasets used for training. Thousands of artists have signed a petition demanding that AI developers cease using copyrighted music without licensing. The lawsuits seek to establish that the use of these recordings violates U.S. copyright law.

At the federal level, legislation has been introduced to protect creators’ rights. Senators Marsha Blackburn (Tennessee) and Chris Coons (Delaware) have sponsored a bipartisan bill that would make it illegal for digital replicas to be created without the creator’s consent. The Senate Judiciary Committee passed the bill unanimously last week, signaling strong congressional support for addressing the issue.

The broader industry implications are significant. AI music‑generation platforms, such as Suno, Udio, and other emerging tools, rely on large datasets of existing recordings to learn musical patterns. The lack of transparency about training data raises questions about compliance with copyright law and the ethical use of artists’ work. As AI models become more sophisticated, the demand for high‑quality training data is expected to grow, potentially increasing pressure on independent creators.

For now, Paddock and his peers are calling on the music community to support independent artists. He urges fans to attend live shows, purchase music, and share work from lesser‑known creators. “Now more than ever is the time to support the artists that you feel like are the most human,” he said.

The situation remains unresolved. Artists continue to rely on public pressure, legal action, and forthcoming legislation to protect their intellectual property. The AI Watchdog database will likely remain a key resource for tracking the use of copyrighted music in AI training, and the outcome of the pending federal bill will shape the future relationship between AI developers and the music industry.