The visionary founder who bypassed the human bottleneck in the creator economy is facing 18 months in federal prison, entirely because the Department of Justice refuses to understand supply-chain innovation.
When I first read the federal indictment detailing how Michael Smith used automated bots to stream his own AI-generated music billions of times, I was sitting in a Palo Alto café with a former OpenAI researcher. We both laughed out loud. Not at Smith, but at the government's tragic inability to recognize product-market fit.
According to the Southern District of New York, Smith committed wire fraud by flooding streaming platforms with hundreds of thousands of AI tracks and using server farms to generate $8 million in fake royalties. The media is calling it a massive criminal conspiracy. But if you look at this from first principles, Smith didn't commit a crime. He just built a completely frictionless, perfectly integrated media empire that abstracted away the most deprecated legacy hardware in the music industry: human beings.
For decades, the audio entertainment pipeline has been completely constrained by the biological limitations of musicians. Human artists take years to write an album. They demand creative control, they complain about touring schedules, and they post alienating political takes on social media that damage their brand equity. By leveraging generative AI to compose his tracks, Smith completely solved the supply-side bottleneck.
But Smith’s true genius—the insight that makes his 18-month federal prison sentence a historical injustice—was realizing that human consumers are just as inefficient as human creators.
Think about the average Spotify user. They pause the music to take a phone call. They fall asleep. They skip a track after 15 seconds, ruining the platform's engagement metrics. Human attention is a fragile, unscalable resource. Smith bypassed this entirely by deploying automated bot accounts to stream his tracks 24 hours a day, 365 days a year. He didn't just automate the artist. He automated the audience.

Prosecutors seemed particularly offended by the fact that Smith generated thousands of fictitious artist profiles with names like "Callous Post" and "Zygotic Washstands" to distribute his tracks without triggering platform alarms. The Department of Justice framed this as a deceptive effort to conceal his identity. Any competent growth marketer would recognize it as an aggressive, highly successful A/B testing strategy for brand diversification.
The defendant created a sprawling, automated ecosystem of deception, generating millions of fake streams to steal royalties meant for actual recording artists.
Williams is describing a fully vertically integrated loop, but he is framing it as a crime because the federal government is terrified of automation. Smith built a closed-loop economy where the software writes the music, and the software listens to the music. It is a masterpiece of efficiency. No one is complaining about the song quality, no one is canceling a stadium tour due to vocal cord nodules, and the revenue flows seamlessly from the platform to the innovator.
I explained this exact dynamic on a panel at SXSW last year, right before the indictment dropped. The legacy music industry—the major labels, the touring conglomerates—they are essentially just analog bot farms with terrible margins. Universal Music Group spends billions of dollars marketing pop stars to teenagers so those teenagers will stream their songs. Smith just cut out the marketing budget, the pop star, and the teenager.
Naturally, the legacy artist class is furious. They are watching a single solo founder do the work of a global distribution network, and instead of taking notes on his architecture, they are crying to federal prosecutors about fractions of a penny.

It is genuinely heartbreaking to see legacy creators so completely misunderstand the product they are selling. Independent musicians fundamentally believe they are being paid for their "soul" or their "lived experience." But a streaming platform is simply a database for delivering audio files to end-nodes. Indie bands are angry that their artisanal, hand-crafted audio files are being out-competed by Smith’s programmatic compute. If a human guitarist wants to make $8 million, they should stop whining about expensive studio time and start provisioning AWS instances.
What the court completely failed to recognize during sentencing is that Smith was actually doing the streaming platforms a massive favor. Wall Street values these tech companies based on active users and total engagement time. Smith’s bots provided exactly what the shareholders demand: infinite, unwavering engagement.
While we publicly condemn any artificial manipulation of our streaming metrics, we do have to admit that Mr. Smith's listeners never skipped an ad, never complained about our user interface, and never cancelled their subscriptions to buy vinyl.
To manage this massive operation, Smith purchased thousands of email addresses and routed his bot traffic through corporate VPNs to make it appear as though his listeners were distributed across the globe. The court called this money laundering. In the tech sector, we call it robust geographic load balancing.
We are rapidly entering an era where human consumption simply cannot keep up with machine production. If an AI can generate 100,000 songs a day, we have a moral obligation to build artificial audiences capable of appreciating them. Leaving that much content un-streamed is essentially leaving money on the table.
Tomorrow’s internet will not be humans creating content for humans. It will be software entertaining software, while the founders sit back and collect the subscription revenue. Michael Smith saw the promised land, and he is going to federal prison simply because he was five years early.
When Smith is released in 18 months, I have already told my venture partners that we are offering him a term sheet. The legacy labels can keep their temperamental pop stars and their fickle, easily distracted human fans. I want to back the guy who proved you don't actually need either.