An $8 Million AI Music Fraud Comes to a Sentence: In the Streaming Royalty Pool, Every Fake Play Steals From Real Musicians

Updated:

On October 6, 2026, the U.S. federal court for the Southern District of New York sentenced Michael Smith of North Carolina to 18 months in prison, 2 years of supervised release, and forfeiture of $8,091,843.64 (over 57 million RMB) in criminal proceeds, after his guilty plea on one count of conspiracy to commit wire fraud. According to the U.S. Department of Justice press release, this is the first criminal case the DOJ has ever brought over AI music streaming fraud to reach sentencing.

Play-counting and royalty distribution mechanisms sit at the heart of this case

Plenty of coverage, particularly in China, compressed this story into "man who used AI to defraud royalty income gets prison time" — accurate as far as it goes, but it misses what makes the case a genuine watershed. The money Smith stole did not come from the platforms or the labels. It was diluted, play by play, out of a proportionally split royalty pool that would otherwise have gone to real musicians. To understand why the industry treats this case as a turning point, start with the machinery of the scheme.

A Fraud Machine That Ran for Seven Years

According to DOJ filings and reporting on the guilty plea, Smith's scheme ran from 2017 to 2024 in roughly three layers.

The first layer was content production. Using AI tools, Smith mass-generated hundreds of thousands of songs. In an email quoted by prosecutors, he said outright that he needed "a TON of songs fast," because streaming platforms' anti-fraud systems monitor anomalous spikes on individual tracks — spread the plays across enough songs, and each one looks "normal."

The second layer was the account matrix. He registered thousands of streaming accounts with fake email addresses and paid for subscriptions with debit cards obtained through unlawful means — paid accounts generate higher royalty rates per play than free ones. At its peak, his bot army ran as many as ten thousand accounts at once.

The third layer was automated playback. Software kept those accounts streaming the AI songs he owned, day after day. The DOJ offered one vivid comparison: in April 2023, U.S. PIRG research indicated Taylor Swift's music received roughly 9.3 million streams on YouTube Music family plans in the previous month; over the same period, Smith's bots generated more than 80.9 million streams of his own music — about 8.7 times as many.

Over seven years, hundreds of thousands of AI songs and billions of cumulative streams earned Smith more than $8 million in royalties. On September 4, 2024, the Southern District of New York indicted him on three counts: wire fraud, conspiracy to commit wire fraud, and money laundering (charging press release). On March 19, 2026, Smith pleaded guilty to the conspiracy count. On October 6, Judge John G. Koeltl handed down the sentence.

Why Fake Plays Are Zero-Sum: The Money in a Pro-Rata Pool Is Fixed

This is the most easily overlooked — and most consequential — part of the case.

Major streaming platforms pay royalties under the so-called pro-rata model: subscription revenue is pooled, and each month the pool is split among rights holders in proportion to their share of total plays. The DOJ's press release unusually devotes a passage to explaining this mechanism and states its consequence plainly: "Royalty payments are made proportionately to musicians and songwriters from a pool of funds. As a result, streaming fraud diverts funds from musicians and songwriters whose songs were legitimately streamed by real consumers."

In other words, the size of the pool is fixed in the short run. Fake plays do not print money out of thin air; they change the split. For every ten thousand plays Smith's bots racked up, all real musicians combined received correspondingly less. The harm is dispersed across millions of rights holders industry-wide — which is exactly what makes it insidious, and, as we will see, what the defense tried to exploit.

A Music Business Worldwide analysis published after the sentencing breaks the arithmetic down further: spread across Smith's hundreds of thousands of songs, an average of just two bot plays per song per day was enough to accumulate $8 million from the pro-rata pool. The piece also notes that Spotify's rule introduced in 2024 — tracks with fewer than 1,000 annual streams no longer earn royalties — would objectively cut the economic lifeline of this "massive catalog, minimal plays" strategy. Had that rule been in place a few years earlier, the scheme's yield would have shrunk dramatically.

Worth noting, too: the platforms' anti-fraud arms race began long before this verdict. In May 2023, Spotify removed tens of thousands of tracks from AI music company Boomy (later partly reinstated). In June 2025, Deezer disclosed that up to 70% of streams of fully AI-generated tracks on its platform were fraudulent, and rolled out an AI-content tagging system. The Smith sentence supplies the missing piece for that battle: criminal deterrence.

The Signal Value of a Landmark Case

Industry bodies responded quickly. The Music Fights Fraud Alliance, IFPI, and RIAA each issued statements after the sentencing, framing it as a warning to the entire streaming-fraud black market. The U.S. Copyright Office has also opened a public inquiry into streaming fraud (MBW report), with legislative and regulatory follow-up in motion.

But perspective matters: Michael Smith is not the whole of streaming fraud, and perhaps not even its typical form. The case's significance is that the complete loop — mass AI production plus bot streaming — was criminally prosecuted for the first time. The industry's everyday reality is different. A study published in February 2025 by CNM (Centre national de la musique) and SNEP, the French recorded-music trade body, estimated that fraudulent plays account for "at least 1–3%" of all streams in France on services like Spotify and Deezer; monitoring firm Beatdapp has previously estimated that around 10% of global streaming activity is fraudulent, on the order of $2 billion in royalties a year. The real black market is dispersed across stream farms and paid play-buying services, and most of it will never see a federal courtroom.

The Sentence: Is 18 Months Harsh or Lenient?

The sentencing-memo tug-of-war before the verdict maps the two sides of the case neatly.

Prosecutors sought at least 46 months. In reporting on their filing, they argued he "stole from hard-working songwriters" and pressed for a heavier sentence to deter would-be imitators.

The defense asked for no prison time — home confinement instead — with a striking core argument: in a pro-rata pool, the money diverted from each individual real musician is negligible, and "no artist suffered any perceptible harm" (MBW report). Judge Koeltl ultimately landed between the two at 18 months, with forfeiture set at the jointly agreed $8,091,843.64.

That "no perceptible harm" defense deserves a closer look, because it exposes the structural dilemma of pro-rata: the loss is real — $8 million cannot appear from nowhere; it necessarily equals the sum diluted away from real musicians industry-wide — but spread across each individual victim, the amounts are too small to prove one by one. The more dispersed the victims, the harder the accountability and recovery. That is precisely why this kind of fraud has been so hard to stamp out through civil means (this is the author's inference based on the case materials). Criminal conviction sidestepped that evidentiary dead end — in itself a technical breakthrough of this judicial path.

For comparison: a Danish court gave a streaming-fraud operator a suspended sentence in 2024, so 18 months here marks a clear escalation; but within the U.S. federal sentencing framework, the theoretical maximum for wire fraud conspiracy runs far higher, and voices inside the industry argue that 18 months buys inadequate deterrence. Whether the punishment matches the harm remains an open question.

Coda: Technology Cut the Cost of Faking — Rules Decide the Payoff

Every ingredient of the Smith scheme — mass AI generation, fake accounts, bot playback, pro-rata distribution — predates 2017. What AI did was drive the cost of the one bottleneck, producing enough countable content, to nearly zero. What actually decides whether such schemes survive is the platforms' royalty rules and anti-fraud thresholds: when 1,000 plays becomes the minimum for royalty participation, and AI-generated content carries mandatory labels, the arbitrage space of the mass-minimal-play strategy contracts.

The 18-month sentence and $8 million forfeiture close the chapter on the first generation of AI music fraud. But as long as the pool remains a pool and the stream remains a countable unit, the battle over the split will not end — the next front may simply replace "fake plays" with something harder to detect (this is the author's inference).

Sources

  1. ITHome: Man in the US sentenced to 18 months and $8m forfeiture for defrauding music royalties with AI (Subo, 2026-10-08) — original article: basic facts of the case and the sentencing
  2. U.S. DOJ: North Carolina Musician Sentenced To 18 Months In Prison For Music Streaming Fraud Aided By Artificial Intelligence (2026-10-06) — official sentencing release: charges, sentence, forfeiture, official explanation of pro-rata distribution, Taylor Swift stream comparison
  3. U.S. DOJ: North Carolina Musician Charged With Music Streaming Fraud Aided By Artificial Intelligence (2024-09-04) — 2024 charging release: the three counts and the "first-of-its-kind" framing
  4. Music Business Worldwide: Man behind $8m AI song and bot streaming fraud is sentenced to 18 months in prison (2026-10-06) — sentencing details, the 46-vs-18-month gap, MFFA/IFPI/RIAA statements, the Danish comparison
  5. Music Business Worldwide: An $8m streaming fraud, two plays a day — the pro-rata math behind the bot scheme (2026-10) — the two-plays-a-day pro-rata arithmetic and the impact of Spotify's 1,000-stream threshold
  6. Music Business Worldwide: Prosecutors want man who pocketed $8m using AI songs imprisoned for at least 46 months (2026-09) — the government's sentencing recommendation and its reasoning
  7. Music Business Worldwide: Man who pocketed $8m asks for no prison time, arguing no artist suffered any perceptible harm (2026-09) — the defense's "no perceptible harm" argument
  8. Music Business Worldwide: Streaming fraud man who pocketed $8m… pleads guilty (2026-03-19) — guilty plea date and charge details
  9. Music Business Worldwide: Deezer rolls out AI tagging system; up to 70% of streams from fully-AI-generated tracks are fraudulent (2025-06) — Deezer's AI tagging system and the 70% fraudulent-stream figure
  10. Music Business Worldwide: Streaming fraud accounts for 'at least 1-3%' of plays in France (2025-02) — the CNM/SNEP study on fraudulent stream share in the French market
  11. Music Business Worldwide: Music streaming fraud inquiry launched by US Copyright Office — the U.S. Copyright Office's streaming-fraud inquiry