Michael Smith pleads guilty to fraud scheme built on AI-generated tracks and fake streams

North Carolina resident Michael Smith has pleaded guilty in a major music streaming fraud case. According to prosecutors, he uploaded large volumes of AI-generated tracks and inflated their play counts with bots, so he could collect royalties from the platforms' shared pool.

How the scheme worked

Investigators believe Smith built a nearly industrial operation. He created thousands of accounts on music services and ran software that played his tracks nonstop. To avoid drawing attention with abnormally high numbers on a single song, the plays were spread across a huge catalog.

This is where AI gave him a key advantage: instead of recording real music, he could quickly churn out hundreds of thousands of compositions suited to mass uploading and equally mass play inflation.

Court materials say the fake traffic went through major platforms such as Amazon Music, Apple Music, Spotify and YouTube Music. In total, the bot accounts generated billions of plays.

For US authorities, this is one of the first truly prominent cases in which generative AI was not merely a contested creative tool but a direct part of financial fraud. Here AI was used not to experiment with music but as a way to scale deception in a manner that would previously have been too slow and expensive.

The money and the scale

The story dragged on for years. When charges were filed in September 2024, prosecutors said the scheme ran from 2017 to 2024 and could bring in more than a million dollars a year. Court documents cited a figure of up to 661,440 plays per day, and the initial estimate of illegally obtained royalties exceeded $10 million.

Under the current plea deal, Smith pleaded guilty to conspiracy to commit wire fraud and agreed to forfeit $8,091,843.64.

• the scheme lasted about seven years
• daily inflation reached 661,440 streams
• the deal provides for forfeiture of $8,091,843.64
• the maximum penalty under the charge is up to five years in prison

The forfeiture figure itself shows this is not a gray area of music marketing, nor typical play inflation to promote a little-known artist. It is a full-fledged scheme to extract money from the royalty distribution system.

The court is due to hand down its sentence on July 29, 2026. Even if the final term falls below the maximum, the case has already become an important precedent: the state is taking the position that inflating plays with AI and bots is not a marketing trick but criminal fraud.

Why the case matters

Streaming economics work so that payouts to musicians come from a shared pool of money in proportion to the number of plays. That means fake streams do not just distort a platform's metrics; they literally take money from real artists, songwriters and rights holders.

For that reason, the victims in this case include not only the services but also musicians whose tracks were listened to by real people. Prosecutors explicitly describe the scheme as a redistribution of royalties from genuine performers to a man who created both the music and the audience artificially.

"The songs and listeners were fake, but the millions stolen were very real," said US

Attorney Jay Clayton.

The case goes beyond a single criminal story because music platforms are already drowning in a wave of AI content. Deezer previously estimated that it receives around 60,000 fully AI-generated tracks every day, and many listeners cannot tell machine-made music from human-made music at all. Against that backdrop, Smith's scheme shows the central risk of the generative era: the problem is not only who wrote a track, but who can automate the production, publication and fake consumption of content in a single loop.

What this means

For streaming services, it is a signal to speed up anti-fraud systems and scrutinize anomalous catalogs more closely, not just suspicious spikes on hits. For the market as a whole, the case matters because the debate over AI in music is shifting from copyright to money and distribution infrastructure.

The question now is: who collects the royalties if both the music and the listeners can be synthesized with almost no human involvement?