He Faked $8 Million in Streams With an AI Bot Army. Every Producer Splitting a Royalty Pool Paid For It.

Streaming royalties are not a per-song tab. Every major platform pools subscription and ad revenue for a given market and month, then splits it pro-rata by share of total streams. That single design choice is why a bot farm in North Carolina and a betting site in New York both landed in the same story this year: fake streams and real streams draw from the same pot, so every fraudulent play quietly resizes what is left for everyone else.

How did one man steal $8 million in streaming royalties?

Michael Smith, 54, pleaded guilty in March 2026 to a single count of conspiracy to commit wire fraud before US District Judge John G. Koeltl, the first criminal case in the US built around AI-generated music streaming fraud. According to prosecutors and reporting on the case, Smith used AI tools to generate hundreds of thousands of songs, then spread them across roughly 1,000 bot accounts that streamed them around the clock on Spotify, Apple Music, Amazon Music and YouTube Music. Spreading the plays across that many accounts, rather than hammering one, was the entire trick: it let the fraud sit under the radar of each platform's anomaly detection for years while racking up billions of streams. Smith agreed to forfeit $8,091,843.64, the royalties he collected, and faces up to five years in prison. His sentencing is scheduled for July 29, 2026.

Spreading fake plays across a thousand quiet accounts, instead of one loud one, is what let a bot farm hide inside a legitimate royalty pool for years.

Why is Spotify cracking down on Kalshi bets the same month?

A second, unrelated case is running on a parallel track. Spotify pulled roughly 500,000 fraudulent streams from a chart-topping track after tracing unusual activity back to betting on Kalshi, a prediction-market platform where users can wager on which songs will chart. That is a new incentive layered on top of an old problem: streaming fraud used to be about royalty theft alone; now a chart position itself can be worth real money on a betting market, giving someone a reason to buy streams even if they never see a cent of royalties. Spotify has responded by tightening its chart-manipulation detection, and the Kalshi episode is being read industry-wide as an early warning that prediction markets will keep finding new angles to game the same vulnerable signal: raw stream counts.

What does this mean for underground and independent producers?

Neither case targeted house or techno specifically, but the mechanics land hardest on exactly the producers who have no cushion. Major-label catalogs and A-list pop acts absorb a diluted royalty pool without much pain; a bedroom producer on a small label splitting a few thousand streams a month feels every fraction of a cent that bot traffic siphons off. The pro-rata pool model means Smith's billions of fake streams did not just steal from the platforms, they mathematically reduced the share available to every legitimate stream that month, dance music included. Detection also cuts both ways: platforms tightening anti-fraud systems to catch bot farms and Kalshi-style gaming can also flag small, unusual-looking accounts and regional spikes that are common for niche electronic scenes, meaning the fix carries its own risk of false positives against the artists it is meant to protect.

Why it matters

A five-year maximum sentence and an $8 million forfeiture are real consequences, but they are a rounding error against how cheap generative tools have made mass fake-song production, and Kalshi shows the next fraud incentive is already live. Expect platforms to keep leaning on stricter detection, which will affect how every small label's streams get counted, not just the fraudsters'.

What we think

This case proves what independent producers have said for years: the royalty pool is porous and the platforms only chase fraud loud enough to make headlines. A thousand-account bot farm ran for years before anyone caught it. The real fix isn't a single prosecution, it's moving off a pure stream-count model that rewards volume over listener intent, because as long as raw plays set the payout, someone will keep finding a cheaper way to fake them.