The $8 Million Wake-Up Call: Why Streaming Platforms Need a Dual Structure for AI Music

Last week, Michael Smith of Cornelius, North Carolina pleaded guilty in federal court to running a scheme that exploited music streaming platforms and diverted royalty payments from artists. He used AI bots and agents to generate hundreds of thousands of songs, and automated accounts to stream them billions of times. He’s been ordered to forfeit $8 million. Prosecutors are calling it the first criminal streaming fraud case in U.S. history.

The fraud was audacious, but it exposed something the industry would rather not discuss: the underlying system was already broken before Smith abused it. He got hit for wire fraud, but click fraud isn’t the only issue here.

The Quieter Theft

Smith’s scheme was illegal because he used bots. But a version of the same economic harm is happening every day, perfectly legally. Deezer alone is receiving more than 60,000 fully AI-generated tracks daily. Every one of those tracks competes for a share of the same royalty pool that human artists depend on. Humans who take risks, and put massive effort into their skills. The skills that in fact are the only reason AI models can exist in the first place.

Streaming royalties aren’t paid per stream at a fixed rate — they’re paid as a fraction of a shared pool. When AI-generated content floods that pool with millions of streams at near-zero production cost, the per-stream value drops for everyone else. A human artist who spent two years writing, recording, and funding an album competes for the same fractions as content that cost someone an API call and a bit of iteration.

That isn’t fraud (as long as the producer is honest about it, anyway), but it isn’t fair either.

A Structural Solution

The fix isn’t to ban AI music — that ship has sailed, and frankly, as someone who creates AI-assisted music myself, I wouldn’t want it to. The fix is structural separation.

Streaming platforms should implement a dual-pool revenue model:

• AI music gets its own clearly labeled category, with transparent disclosure at upload, enforced through distributor attestation and liability for misclassification.
• Subscribers choose their experience: a standard subscription (human artists only), an opt-in AI tier, or AI-only if that’s their preference.
• Revenue pools remain separate. AI streams fund AI artist royalties. Human streams fund human artist royalties. The two never dilute each other.

The Imperfect Enforcement Objection

Someone will point out that this system can be gamed. Artists could mislabel AI content as human to access the larger royalty pool. True. But copyright attestation — the system we already use — is also gameable and regularly gamed. We accept that imperfect enforcement is better than no enforcement. The same logic applies here. Distributor-level attestation with legal liability for misclassification is about as enforceable as anything else in this space.

Why I’m Arguing This

I produce AI-assisted music. My work sits in the gray area this proposal would need to define. I have a direct financial interest in AI music being treated fairly on streaming platforms.

I’m arguing for this anyway because the current trajectory isn’t sustainable for anyone. If AI floods the royalty pool unchecked, human artists exit the streaming ecosystem, platform catalogs degrade, and the creative ecosystem that AI tools are built on top of hollows out. That’s bad for everyone, including AI music creators. With ALL AI, we need to think long term. Short term thinking is inherently destructive (another article on that coming soon.)

The Michael Smith case got prosecuted because he used bots. The broader structural problem won’t get fixed by prosecution. It requires platforms to act — and pressure from artists, listeners, and yes, AI music creators, to make them do it.

The tools exist. The model is straightforward. The only missing ingredient is will.