The Scheme That Ran for Years Before Anyone Was Charged
Michael Smith was not a musician. He was a suburban dad from North Carolina who owned a chain of urgent-care clinics and figured out a way to turn AI-generated audio slop into a personal ATM — at the direct expense of every legitimate artist on Spotify, Apple Music, Amazon Music, and YouTube Music.
The Southern District of New York announced Smith had pleaded guilty to conspiracy to commit wire fraud. The charges stem from a scheme in which he generated hundreds of thousands of songs using AI tools, then deployed networks of bots to stream those tracks billions of times, simulating real listener activity. The total take: over $8 million in royalties. He's now agreed to forfeit that entire sum and faces up to five years in prison, with sentencing scheduled for July 29.
This is the first major criminal conviction tied specifically to AI-generated music fraud — and it almost certainly won't be the last.
How the Math Actually Worked
Streaming royalties aren't a fixed per-play fee. They come from a shared pool: platforms collect subscription and ad revenue, then divide it proportionally based on how many streams each track receives relative to total platform-wide listening. That architecture means every fake stream is a direct theft from every legitimate stream happening at the same time.
Smith understood this perfectly. According to a Rolling Stone investigation, he operated 1,040 accounts across platforms, each one cycling through roughly 636 songs per day. By his own estimates, that generated around $3,300 daily — over $1.2 million a year — with bots doing all the "listening."
The scheme also had a key design principle: distribute streams thinly enough across enough tracks to avoid triggering fraud detection. Hundreds of thousands of songs, each pulling a modest stream count, adds up to billions of plays while staying below any single alert threshold. It was a patient, methodical operation.
Although the songs and listeners were fake, the millions of dollars Smith stole was real.
— Jay Clayton, US Attorney, Southern District of New York
This Is Bigger Than One Guy
Smith's conviction matters as a precedent, but the underlying problem it exposes is structural and ongoing.
Streaming platforms have known about royalty fraud for years. Spotify claims to invest heavily in detecting and removing artificial streams, and has updated its policies to ban AI impersonation and require disclosure labels. But the gap between policy and enforcement has been wide enough for operations like Smith's to run for years at scale before criminal charges materialized.
The real damage isn't just financial — it's competitive. Royalties come from a fixed pool. Every dollar Smith extracted was a dollar not paid to an independent artist with real listeners. At the same time, platforms' recommendation algorithms are volume-sensitive; a flood of AI-generated tracks suppresses the discoverability of genuine music by sheer numerical weight. Small and mid-tier artists are effectively fighting for oxygen in a room that someone else is quietly vacuuming.
This puts pressure on Spotify and Apple Music in particular, because their royalty distribution models are uniquely vulnerable to this kind of manipulation. A per-stream fixed rate would eliminate the zero-sum dynamic. Both companies have resisted that model, and cases like Smith's illustrate the cost of that choice for the artists they depend on.
The Broader AI Music Problem
Smith's case sits inside a much messier landscape. The fraud here was clean in legal terms — fake accounts, fake streams, clear intent — but AI's disruption of the music industry extends into murkier territory that's harder to prosecute.
Entirely AI-generated acts have accumulated millions of legitimate streams from real listeners who simply don't know or don't care that no human made the music. High-profile artists like Drake have dealt with deepfaked vocals appearing on viral tracks they had no involvement with. And platforms are now flooded with low-effort AI compositions that game search and recommendation systems through keyword stuffing and volume.
The Smith case is the easy version of this problem — a clear bad actor doing clearly illegal things. The harder version is the ambient erosion of the music economy by AI content that's technically legal but economically devastating to working musicians.
What This Means
- For streaming platforms: A guilty plea is useful PR, but it doesn't fix the architecture. Platforms need to explain publicly how their fraud detection failed to catch billions of fake streams for years, and what's structurally different now. Vague claims about "investing in detection" aren't enough anymore.
- For independent artists: The conviction is a signal that prosecutors are paying attention — but don't mistake one case for a solved problem. The royalty pool dilution from AI-generated content continues regardless of whether the streams are fraudulent in the legal sense.
- For music rights holders and labels: The royalty pool model is increasingly indefensible. This case is a preview of the lobbying ammunition artists' groups will use to push for per-stream fixed rates or AI-content exclusion from royalty pools entirely.
- For AI developers building generative audio tools: The tools themselves aren't illegal. How they're used is. But expect increasing regulatory scrutiny of platforms that enable mass music generation without any safeguards against downstream fraud.
- For prosecutors: Smith's case establishes that wire fraud statutes apply cleanly to this category of AI-assisted royalty theft. That's a template other DAs can work from.
The Bigger Picture
There's a version of this story that's almost absurdly mundane: a middle-aged clinic owner figured out a side hustle, automated it, and ran it until the feds showed up. But the implications are anything but mundane.
The music industry is now facing a structural challenge that copyright law, platform policy, and fraud prosecution are all scrambling to keep up with simultaneously. Smith's scheme worked because the incentives were misaligned at every level — AI made content creation trivially cheap, royalty pools made fake streams financially rewarding, and detection systems were too slow to catch the pattern before hundreds of millions of dollars in damage had been done.
One conviction doesn't fix any of that. It just proves the game can eventually have consequences.
