Entries / ·

Bot accounts and a contracted supply of AI songs used to take $8m in streaming royalties

Michael Smith, 54, of Cornelius, North Carolina, obtained more than $8 million in streaming royalties between 2017 and 2024 by running roughly 10,000 bot accounts against music he owned on Spotify, Apple Music, Amazon Music and YouTube Music. From 2018 he bought that catalogue — hundreds of thousands of AI-generated tracks — from an AI music company executive under a monthly supply contract. He pleaded guilty to conspiracy to commit wire fraud in March 2026.
Mechanism — Platform detection looked for implausible play counts on individual tracks. Volume was the binding constraint on evading it, and a contracted supply of AI songs removed that constraint — the same revenue, spread thin enough per track to sit inside normal variation.

Correcting a widespread description

Nearly all coverage of this case describes Smith as having created hundreds of thousands of songs with artificial intelligence. The indictment says something different: he bought them.

The misdescription does not originate with the press. It is in the United States Attorney’s own statement announcing the plea, which says Smith generated the songs using artificial intelligence — contradicting the charging document filed by the same office eighteen months earlier. Trade coverage then reproduced that framing while linking to the indictment in the same paragraph. The Help Net Security piece listed below is included as an example of that, not as a source for any fact in this entry.

The supplier was the chief executive of an AI music company, charged in the indictment as a coconspirator. Under a Master Services Agreement dated 1 February 2019, that company undertook to deliver between one thousand and ten thousand songs a month, with full intellectual property rights passing to Smith. Smith paid the greater of two thousand dollars a month or fifteen per cent of the streaming revenue those songs earned. A music promoter, charged as a further coconspirator, and the supplier each took ten per cent of the proceeds.

That distinction matters for anyone trying to understand how this category of fraud scales. It was not one person with a generation tool. It was a procurement relationship with a contracted monthly volume, a revenue share, and a supplier whose product quality improved over the life of the contract — by August 2020 the supplier was reporting substantially better audio and newly added vocal generation.

The constraint AI actually removed

Smith had been running bot accounts since 2017, initially against his own small catalogue and then against a publicist’s. He also tried selling fraudulent streams to other musicians as a service. Neither produced the volume the scheme needed.

The reason is a detection threshold. Platforms watched for individual tracks with implausible play counts, so a billion streams concentrated on one song was conspicuous while the same billion spread across tens of thousands of songs was not. Catalogue size, not bot capacity, was the limiting factor.

This is unusually well evidenced. The indictment quotes Smith’s own emails to coconspirators across late 2018 and 2019 stating the logic directly: that he needed a large quantity of content carrying small numbers of streams each, that he needed songs quickly in order to work around the anti-fraud systems then being deployed, and that without enough catalogue he would overrun individual songs and lose them. In June 2019 he asked his supplier for a further ten thousand songs specifically so the streams could be spread wider.

Operational detail

Roughly ten thousand bot accounts at peak, registered through bulk-purchased email addresses in fictitious names, with over a thousand running simultaneously. Family plans were used because they were the cheapest way to hold multiple accounts. VPNs concealed that the accounts all operated from one residence. Payment was routed through a debit-card service that Smith told he was issuing cards to company employees, funded with $1.3 million in fraud proceeds.

The supplied audio files arrived with randomised filenames. Smith generated plausible song and artist names to attach to them — the disguise was in the metadata rather than the audio.

Three figures, not one

In a February 2024 email Smith claimed over four billion streams and twelve million dollars in royalties since 2019. The indictment alleges more than ten million. The plea carries a forfeiture of $8,091,843.64.

These are not revisions of a single estimate. The first is an offender’s boast, the second a prosecutor’s allegation across three counts, the third a negotiated forfeiture attached to one count. Coverage that presents them interchangeably, or that pairs the largest figure with the guilty plea, is conflating three different quantities.

The provenance lie

When the Mechanical Licensing Collective halted payments and challenged him in 2023, Smith and his representatives denied stream manipulation and asserted that the works were human-authored rather than computer-generated. The false claim was about AI provenance specifically, which is worth noting as platforms and collecting societies move toward mandatory AI disclosure — the enforcement question is not only whether disclosure is required but whether a denial is independently actionable.

Enforcement

Charged 4 September 2024 on three counts — wire fraud conspiracy, wire fraud, and money laundering conspiracy — each carrying a maximum of twenty years. Pleaded guilty 19 March 2026 to a single count of conspiracy to commit wire fraud, maximum five years, forfeiture agreed at $8,091,843.64. Sentencing was listed for 29 July 2026 before Judge John G. Koeltl.

Open — sentencing outcome not confirmed in this record. The listed date has passed. Also unresolved: the indictment identifies the AI music company executive and the promoter as coconspirators, and no public charging document against either has been located.


Sources

  1. P U.S. District Court, SDNY — Sealed Indictment, United States v. Michael Smith, 24 Cr. 504 retrieved 2026-09-14
  2. P U.S. Attorney's Office, SDNY — North Carolina Man Pleads Guilty To Music Streaming Fraud Aided By Artificial Intelligence retrieved 2026-09-14
  3. S Bloomberg Law — AI Music Maker Who Faked Streams Pleads Guilty on Fraud Count retrieved 2026-09-14
  4. T Help Net Security — Fake AI songs streamed billions of times, netting fraudster $10 million retrieved 2026-09-14

Search used — site:justice.gov "artificial intelligence" streaming fraud royalties

Entry last updated 14 September 2026. Tier definitions and status meanings are set out in the methodology.