AI song fraud siphons $8M from music royalty investors

A North Carolina man's AI-powered streaming fraud exposed a structural weakness in the $8 billion market for music royalty-backed bonds.

Michael Smith used artificial intelligence to generate hundreds of thousands of songs and deployed bot accounts to stream them billions of times, siphoning more than $8 million in royalties over seven years — a scheme that prosecutors called the first criminal streaming fraud case in the United States.

"The defendant exploited the very mechanics of how streaming royalties are calculated," Damian Williams, US Attorney for the Southern District of New York, said in a statement announcing the guilty plea.

Smith ran as many as 10,000 bot accounts, generating an estimated 661,440 streams a day, according to prosecutors. He agreed to forfeit $8,091,843.64 and faces sentencing July 29. The Mechanical Licensing Collective, which distributes streaming royalties to American songwriters, flagged the anomaly early and challenged Smith's representatives, preventing the diversion of those particular royalties.

The case lands as Wall Street has turned music royalties into an $8 billion bond market, with firms including Blackstone-backed Hipgnosis packaging song catalogues into investment-grade securities. The entire edifice rests on a single assumption — that the royalty pool remains predictable. AI-generated tracks now threaten that assumption.

Deezer, a streaming service, reported in April that AI-generated tracks make up 44% of all music uploaded to it each day, roughly 75,000 synthetic songs, up from about 10,000 a day a year earlier — a sevenfold jump. While actual consumption of AI music sits at just 1% to 3% of total streams, and 85% of AI track streams are flagged as fraudulent, the supply pressure on the royalty pool is building.

The fraud mechanics matter for investors. Streaming services do not pay a fixed sum per play. Instead, they pool monthly subscription and advertising revenue and divide it across all streams. More streams chasing the same pool means each stream is worth less. A catalogue can attract the same listeners and still earn less because more songs are splitting the same money.

How the Royalty Pool Works

The Department of Justice spelled out the math in the Smith case: royalties are "made proportionately to musicians and songwriters from a pool of funds." That pool is broadly fixed in any given month. If AI floods the pool with supply — lawful or fraudulent — the share attached to any one human catalogue is squeezed. The quality these assets were sold on, dependable and predictable cash flow, is the quality a supply shock to the royalty pool erodes.

The risk is not theoretical. Hipgnosis sold a $1.47 billion bond secured on a catalogue of roughly 45,000 titles in 2024, including songs by the Red Hot Chili Peppers, Neil Young, Journey, Shakira and Bon Jovi. The Lyra bond carried an A- rating from a ratings agency, comfortably inside investment grade. Primary Wave closed a music fund at $2.225 billion in April. More than $8 billion of music-backed bonds have been issued since 2020.

A further question hangs over the arrangement: US royalty rates are set through a government process called the Copyright Royalty Board, and it has not yet settled whether AI-generated tracks are entitled to the compulsory license that human-made songs draw on. If the machines diluting the pool turn out to have a contested claim on it, the rules governing who gets paid could shift under the funds that have already bought in.

What Investors Should Watch

For income-seeking investors, music royalties remain genuinely independent of equity markets. But the headline yield is only the start of the homework. Investors should ask whether a fund's catalogue is human and verifiable, how the manager models future royalties, and whether the platforms' fraud detection — Deezer's detector, Spotify's fraud measures — is keeping pace with the flood. The value of the holding depends on machines on the other side winning a race most investors never knew was being run.

This article is for informational purposes only and does not constitute investment advice.