Michael Smith gets 18 months, and the Copyright Office finally asks how streaming fraud works

Every fake stream is paid for by a real artist. That is the plain arithmetic behind the case of Michael Smith, whom the Justice Department calls the first American criminally charged with AI-assisted streaming fraud. On 6 Oct 2026 he was sentenced to 18 months in prison. The next day, the US Copyright Office opened a public inquiry into how streaming fraud works and what to do about it. Artists now have a formal channel to say what this costs them, and a deadline: 23 Nov 2026.

What did Michael Smith do?

Smith, a 54-year-old musician from North Carolina, was sentenced to 18 months in prison and ordered to forfeit $8,091,843.64, Ars Technica reports, citing the Justice Department. BleepingComputer adds two years of supervised release. He had pleaded guilty in March 2026 to one count of conspiracy to commit wire fraud in the Southern District of New York, where Judge John G. Koeltl set the 6 Oct sentencing, according to the Copyright Office's own notice.

The mechanics were simple and patient. From 2017, Smith used fake email addresses and fraudulently obtained debit cards to open thousands of accounts on Spotify, Apple Music, Amazon Music and YouTube Music, then ran software that kept those accounts streaming songs he claimed to own. By his 2024 arrest he had hundreds of thousands of AI-generated tracks, and he spread billions of plays across them so that no single song looked suspicious, per Ars. BleepingComputer reports that the bots connected through VPNs, and that in a 2018 email Smith wrote that he needed a large number of songs quickly to get around the platforms' anti-fraud rules. That is the point of the AI: not to make hits, but to make enough tracks that the fraud disappears into the noise.

Ars reports it ran for seven years without detection. Smith's lawyers asked for probation, arguing he was being punished for "the ills of an industry" and had not actually kept $8 million. Prosecutors argued that he used increasingly deceptive tactics and that the sentence had to deter others, Ars reports.

Who actually lost the money?

Mostly other artists. As the Copyright Office's notice explains, services generally set aside a share of revenue as a royalty pool and divide it by plays ("pro-rata"). A fraudulent play doesn't create new money; it takes a slice of a fixed pie. The notice names the primary harm as dilution of that pool, and lists secondary ones: distorted playlist and algorithm placement, and corrupted data that artists and venues use to decide when and where to tour.

That is why Smith could run for seven years. His take was spread across the royalty pools of four services, so no single artist would have seen a line on a statement that said "stolen".

What is the Copyright Office asking?

The notice of inquiry, published in the Federal Register on 7 Oct 2026, answers a 21 May 2026 letter from Rep. Scott Fitzgerald asking the Office to examine how common streaming fraud is and how AI feeds it. Fitzgerald leaned on Deezer's numbers. In April Deezer said roughly 44% of its daily uploads were fully AI-generated, and that 85% of the streams on those tracks were detected as fraudulent and demonetised. Deezer also says AI tracks are only 1–3% of its total streams, so the problem is concentrated, not everywhere.

The Office's ten questions include:

• How fraud is executed, and what data exists on bot farms, click farms and playlist manipulation.
• Which industry policies or practices may be "inadvertently contributing" to fraud.
• How fraud relates to AI-generated music and to royalty licensing.
• How well trusted-distributor rules, customer verification, royalty clawback, fines and content removal actually work.
• Whether a shared database of known fraud, or labelling content ineligible for statutory royalties, would help.
• Voluntary, statutory or regulatory fixes, including no-fault injunctive relief.

The notice also records that some fraudsters simply move from distributor to distributor once caught, and that others have proposed replacing pro-rata with an artist- or user-centric payout. Initial comments are due by 11:59pm US Eastern on 23 Nov 2026, replies by 21 Dec 2026, through regulations.gov; instructions are on the Copyright Office page. The Music Network has a good summary of the scope.

What we think

Read the dates closely. The notice was signed on 5 Oct, a day before the sentencing. This is not Washington reacting to one case. It is a slow policy process, requested in May, that happens to arrive just as the first prosecution of its kind shows how small the criminal lever is.

And it is small. Eighteen months, after seven undetected years, for one person. Smith's own case shows why prosecution can't carry this alone: generating tracks is now nearly free, the stolen money is spread thin enough that no victim notices, and only the rare case that reaches a courtroom gets punished. Jail time is not a pricing signal anyone running bots at scale will feel.

The useful question in the inquiry is the second one: which industry practices make fraud pay. Artists should answer it specifically, because the honest answer points at the system, not at one man:

• Recovered money should go back to the artists it came from. The notice asks about clawback and fines, but says nothing about where that money goes. If a service fines a distributor or claws back fraudulent royalties, that money came out of the pool and should be returned to it, with a public report of how much.
• Distributors are the choke point. Fraud enters through upload accounts. A distributor that onboards anyone, and a system that lets a banned uploader reappear at the next distributor, are part of the problem. A shared fraud database and real know-your-customer checks are reasonable asks, as long as there is a fast appeal for artists wrongly flagged.
• Test user-centric payouts on the fraud problem, not as an ideology. Under user-centric, a bot account's plays can only redirect that account's own subscription fee, not a share of everyone else's. That doesn't stop fraud funded with stolen cards, but it caps how much a bot farm can drain from real artists. That is our reading, not an established result, and it is exactly the kind of thing the inquiry should ask services to model with real data.

If you're an independent artist, a short comment describing your own experience (unexplained stream spikes, takedowns over "artificial streaming" you didn't buy, money vanishing after a distributor flag) is evidence Congress doesn't currently have. File it before 23 Nov.

Sources

• U.S. Copyright Office, Music Streaming Fraud: Notice of Inquiry, 91 FR 64188 (7 Oct 2026)
• U.S. Copyright Office, Music Streaming Fraud (comment page)
• Rep. Scott Fitzgerald, letter to the Register of Copyrights (21 May 2026)
• Deezer Newsroom, AI-generated tracks now represent 44% of all new uploaded music (20 Apr 2026)
• Ars Technica, Fraudster jailed for using 10K bots and AI songs to outstream Taylor Swift (7 Oct 2026)
• BleepingComputer, Musician sent to prison for streaming fraud using AI bots (7 Oct 2026)
• The Music Network, US Copyright Office Opens Music Streaming Fraud Inquiry (8 Oct 2026)