AI & Music

18 Months for the $8 Million AI Streaming Fraud: What the Case Shows

By the Starchild Music team · October 10, 2026 · 6 min read

Rows of identical glowing speakers in a dark server room, each playing to an empty chair

The short answer

On October 6, 2026, a Manhattan federal judge sentenced Michael Smith to 18 months in prison for a wire fraud conspiracy built on AI-generated songs and bot streams that collected about $8 million in royalties. He must also forfeit $8,091,843.64. Prosecutors had asked for at least 46 months, the defense for no prison time, and industry groups had told the judge that streaming fraud persists because it looks low-risk.

On Tuesday, October 6, 2026, US District Judge John G. Koeltl sentenced Michael Smith, 54, of Cornelius, North Carolina, to 18 months in prison in Manhattan federal court. It was the government's first criminal prosecution centered on artificially inflated music streaming. Smith pleaded guilty in March 2026 to one count of conspiracy to commit wire fraud. According to the court record, he used AI to generate a very large catalog of songs and used automated accounts to stream them, collecting about $8.09 million in royalties between 2017 and 2024.

The sentence

Judge Koeltl imposed 18 months in prison followed by two years of supervised release, and ordered Smith to forfeit $8,091,843.64. The term falls well below the federal guidelines range of 46 to 57 months, below the at least 46 months prosecutors asked for, and below the roughly 24 months the probation office had suggested. The published reporting on the hearing does not give the judge's reasons for going below the guidelines.

After the hearing, the US Attorney's office said, as reported by Music Business Worldwide, that "by flooding music streaming platforms with automated bots in the place of consumers, and fake songs in the place of creativity, Smith robbed millions in royalty payments from genuine artists and their fans." The RIAA said it appreciated "the Court recognizing the significance of this harmful conduct and the ill-gotten gains that come from the pockets of real artists."

What each side asked for

In a filing made public in late September, prosecutors had asked the court to impose "a substantial custodial sentence of at least 46 months imprisonment," the bottom of the federal guidelines range of 46 to 57 months, along with forfeiture of the roughly $8 million he collected. Digital Music News reported that the probation office had suggested a shorter term of about 24 months.

Smith's lawyers argued for no prison time. As reported by Digital Music News and Complete Music Update, the defense described the harm to any individual artist as insignificant, characterized the crime as effectively victimless, and argued that the press coverage the case has already received is deterrent enough.

Prosecutors rejected that framing. In their filing they called Smith "the architect of, and led, a sophisticated and multi-faceted fraud scheme" aimed at money owed to "hard-working and often underpaid" musicians, and warned that leniency would signal that streaming fraud carries little real risk.

The industry's argument to the judge

Two industry bodies weighed in. The Music Fights Fraud Alliance, a nonprofit founded in 2023 whose more than 40 members include Spotify, Amazon Music, distributors, publishers and collecting societies, submitted a letter the Department of Justice then cited in its own filing. Its central point was economic: "Fraudulently generated streams reduce the per-stream royalty rate for all legitimate creators because streaming royalties are drawn from a finite, shared royalty pool." Complete Music Update reported that the alliance told the court fraud is fuelled by "a perceived lack of a legal risk," which is why it wants the sentence to carry a deterrent message.

The Mechanical Licensing Collective, which administers US mechanical royalties for streaming, also filed. According to Digital Music News, the MLC told the court that Smith denied manipulating plays or using AI-generated content after it suspended his payments.

How the scheme worked, according to the record

Music Business Worldwide worked through the numbers in the court record. The setup it describes ran 52 cloud accounts with 20 bot accounts each, about 1,040 bots in total, each able to play roughly 636 songs per day. Smith spread those plays across hundreds of thousands of AI-generated tracks. In his own words, quoted in the filings, the plan required "a TON of content with small amounts of Streams" so that no single track drew attention. MBW's arithmetic puts the result at roughly two plays per track per day.

The pattern was hard to see track by track and easier to see in aggregate. According to MBW's review of the record, the MLC flagged Smith's Amazon activity from 2021 and 2022 as statistically improbable, and his catalog showed unusual concentrations of plays on YouTube Music family plans.

Why pro-rata pools make this everyone's problem

The filings put a spotlight on a structural fact many artists never see. Most major streaming services pay rights holders from a monthly pool divided by share of total streams. A fake stream leaves the platform's total payout unchanged and takes a sliver from every legitimate rights holder in the pool. That is also why, as MBW notes, prosecutors did not seek restitution: the losses are spread too thinly across too many people to produce a list of identifiable victims.

MBW also pointed to a rule change that would have undercut the strategy: Spotify's April 2024 policy requiring a track to reach 1,000 streams in 12 months, along with a minimum number of unique listeners, before it earns recording royalties. A catalog designed for two plays a day falls far below that line. MBW noted the threshold does not yet cover publishing royalties, which flow through other pipes.

Our analysis: what the case tells working musicians

The sentence is lighter than prosecutors and the industry wanted, but it is still prison time, in the first case of its kind, and it strips Smith of everything he collected. The record also settles one point. Cheap generation plus automated listening is a direct transfer out of the shared pool that pays songwriters and performers. The same dynamic sits behind the platform data we covered in Deezer's AI upload figures, where a large share of streams on fully AI tracks is flagged as fraudulent.

For independent artists, the practical lessons are modest but real: register works accurately with your PRO and the MLC so legitimate plays are matched to you, watch your own analytics for unexplained spikes that could trigger fraud holds, and favor distributors that publish their fraud policies. At Starchild, the catalog is built on songs written by people, every song carries an AI-transparency label, and the style models used to produce alternate versions are licensed, with revenue shared with the artists behind them. That kind of documented provenance is the opposite of an anonymous catalog built to be streamed by machines.

Sources