An AI Music Fraudster Pleaded Guilty to $8,091,843 in Fake Royalties—Here’s the Counterintuitive Math That Can Still Pay Real Artists Less
The DOJ says one man used AI-made tracks plus bot “listeners” to siphon millions from streaming’s pooled payout system—shrinking everyone else’s share without changing the “rate.”
By TheMurrow Editorial
April 19, 2026
Key Points
- 1Track the case: DOJ says Michael Smith used AI songs and bots to divert $8,091,843.64 from streaming royalty pools.
- 2Understand the math: pro‑rata payouts divide a fixed revenue pool—fake streams inflate the denominator and shrink real artists’ shares.
- 3See the shift: this is wire-fraud enforcement, not a copyright fight—AI matters mainly because it enables industrial scale and camouflage.
A song can be a three-minute confession, a summer memory, a hard-earned craft. In federal court this spring, prosecutors described another kind of song: one produced in industrial quantities, uploaded in bulk, and played not by people but by automated accounts—until it generated millions in royalties.
Michael Smith, a 52-year-old from Cornelius, North Carolina, pleaded guilty to conspiracy to commit wire fraud, according to the U.S. Department of Justice. The government says Smith built a music-streaming scheme that ran for years, leaning on AI-generated tracks and “bot” listeners to manufacture demand at a scale human artists cannot match. The DOJ frames the harm plainly: the scheme diverted royalty funds away from legitimate musicians and songwriters whose work was actually streamed by real listeners. (DOJ, Southern District of New York)
$8,091,843.64
Forfeiture amount listed in the plea agreement, after prosecutors said the scheme fraudulently obtained more than $8 million in royalties. (DOJ)
The money trail is unusually clear for a case involving the hazy economics of streaming. The DOJ says Smith fraudulently obtained more than $8 million in royalties, and the plea agreement includes a forfeiture amount of $8,091,843.64. Sentencing is scheduled for July 2026, and the charge carries a maximum penalty of five years in prison.
The case matters not because it is the first time someone tried to cheat streaming platforms—fraud is as old as payouts—but because it shows how AI can function as a force multiplier inside a familiar crime. Not a philosophical debate about authorship. Not an abstract fight over training data. A practical question with immediate stakes: what happens to a royalty system when songs become cheap to generate and “listeners” become cheap to simulate?
“When royalties are pooled, fake listening doesn’t just create fake earnings—it can shrink everyone else’s share.”
— — TheMurrow
The guilty plea: what the DOJ says happened—and why it’s different
The defendant is Michael Smith of Cornelius, North Carolina. In the DOJ’s account, his scheme relied on two moving parts working together: a gigantic supply of tracks and an equally gigantic demand signal. Prosecutors say Smith generated “hundreds of thousands” of AI-made songs, uploaded or distributed them, and then used automated bot accounts to stream them billions of times. (DOJ)
Scale is the point. Traditional streaming fraud might focus on a handful of tracks repeatedly looped. The government’s description suggests something closer to an assembly line: large catalog, automated playback, continuous churn. That combination can make detection harder, because the activity resembles the long tail of streaming—millions of songs, each with small audiences—except here the “audiences” are synthetic.
The DOJ’s dollar figures offer a concrete window into the scheme’s size. The press release says Smith “fraudulently obtain[ed] more than $8 million in royalties.” The plea agreement adds specificity: Smith agreed to forfeit $8,091,843.64. Earlier reporting around the September 2024 charge referenced a higher figure—“more than $10 million”—a reminder that charging-stage allegations and plea-stage stipulations can differ as evidence is tested and negotiated. (Forbes; DOJ)
The alleged conduct spans 2017 through 2024, according to DOJ-aligned coverage and major reporting. That length matters: it implies a scheme that adapted across multiple eras of streaming, from playlist culture to the current boom in generative audio. It also means a long period of potential dilution—years in which legitimate artists may have competed, unknowingly, against manufactured activity in the same payout pool.
Scale is the point. Traditional streaming fraud might focus on a handful of tracks repeatedly looped. The government’s description suggests something closer to an assembly line: large catalog, automated playback, continuous churn. That combination can make detection harder, because the activity resembles the long tail of streaming—millions of songs, each with small audiences—except here the “audiences” are synthetic.
The DOJ’s dollar figures offer a concrete window into the scheme’s size. The press release says Smith “fraudulently obtain[ed] more than $8 million in royalties.” The plea agreement adds specificity: Smith agreed to forfeit $8,091,843.64. Earlier reporting around the September 2024 charge referenced a higher figure—“more than $10 million”—a reminder that charging-stage allegations and plea-stage stipulations can differ as evidence is tested and negotiated. (Forbes; DOJ)
The alleged conduct spans 2017 through 2024, according to DOJ-aligned coverage and major reporting. That length matters: it implies a scheme that adapted across multiple eras of streaming, from playlist culture to the current boom in generative audio. It also means a long period of potential dilution—years in which legitimate artists may have competed, unknowingly, against manufactured activity in the same payout pool.
2017–2024
The span of alleged activity in DOJ-aligned reporting and major coverage, implying years of potential royalty-pool dilution.
The legal frame: wire fraud, not copyright
One reason the case has drawn attention is what it is not. The plea is to conspiracy to commit wire fraud, not a headline-grabbing copyright showdown over whether AI “copied” anyone’s sound. The government’s narrative focuses on deception: the misrepresentation of genuine listening activity to secure money.
That distinction signals something important for creators and platforms. Copyright law is slow, technical, and often ambiguous around new technology. Fraud law is older and clearer: if you lie to get paid, the lie matters more than the tool used to craft it.
That distinction signals something important for creators and platforms. Copyright law is slow, technical, and often ambiguous around new technology. Fraud law is older and clearer: if you lie to get paid, the lie matters more than the tool used to craft it.
“The most consequential AI cases may not be about inspiration or imitation, but about industrialized deception.”
— — TheMurrow
The streaming math most listeners never see: why fake plays can cut real pay
Many readers still talk about streaming as if every play has a fixed sticker price—so many cents per stream. Policy documentation tells a different story. Under the dominant model used across major services, royalties are paid from a pro‑rata (“market-centric”) pool: platforms aggregate net distributable revenue and then pay rightsholders based on their share of total streams during a period. (UK Parliament committee documentation)
The simplified logic looks like this:
- (Net distributable revenue ÷ total streams) × your stream share (policy description)
- Your payout depends on:
1) the size of the revenue pool, and
2) the denominator—how many total streams are counted.
That denominator is where fraud becomes more than a platform problem; it becomes an artist problem. If someone injects a vast number of fake streams without adding real subscription revenue or ad value, total counted streams rise. A larger denominator can reduce the “effective” value of a legitimate stream, while the fraudster captures a slice of the same pool.
The DOJ states the harm in human terms: streaming fraud diverts funds from musicians and songwriters whose songs were legitimately streamed by real consumers. (DOJ) That isn’t a rhetorical flourish. Under pro‑rata mechanics, diversion can be built into the arithmetic.
The simplified logic looks like this:
- (Net distributable revenue ÷ total streams) × your stream share (policy description)
- Your payout depends on:
1) the size of the revenue pool, and
2) the denominator—how many total streams are counted.
That denominator is where fraud becomes more than a platform problem; it becomes an artist problem. If someone injects a vast number of fake streams without adding real subscription revenue or ad value, total counted streams rise. A larger denominator can reduce the “effective” value of a legitimate stream, while the fraudster captures a slice of the same pool.
The DOJ states the harm in human terms: streaming fraud diverts funds from musicians and songwriters whose songs were legitimately streamed by real consumers. (DOJ) That isn’t a rhetorical flourish. Under pro‑rata mechanics, diversion can be built into the arithmetic.
The counterintuitive part: “rates” can look steady while shares shrink
Imagine the pool is stable for a month—subscription dollars and ad revenue don’t change much. If total counted streams increase due to automated activity, then the platform is dividing roughly the same pie into more slices. The fraudster’s activity claims slices that would otherwise belong to real songs with real listeners.
People resist this idea because platforms often communicate payouts in per-stream language. Yet pro‑rata systems are fundamentally share-based. When the system is share-based, “fake demand” doesn’t merely mint fake money; it can reallocate real money.
People resist this idea because platforms often communicate payouts in per-stream language. Yet pro‑rata systems are fundamentally share-based. When the system is share-based, “fake demand” doesn’t merely mint fake money; it can reallocate real money.
“In a pro‑rata world, the crime isn’t only stealing from a company—it’s skimming from every artist in the pool.”
— — TheMurrow
Why AI mattered here: not creativity, but scale and camouflage
The DOJ’s core allegation is not that AI music is inherently unlawful. It’s that AI made the scheme easier to run. Generating “hundreds of thousands” of tracks by traditional means would require time, collaborators, studio costs, and—most importantly—scarcity. AI collapses those constraints.
Two features of AI-generated catalogs matter in the government’s telling:
- Volume: a massive number of tracks increases the surface area of the fraud.
- Variety: a large, varied catalog can avoid the obvious red flags of looping a small set of songs.
A single track that receives an implausible number of plays is easy to flag. A large catalog each receiving a plausible trickle is harder—especially when the long tail of legitimate streaming already contains countless niche tracks with modest play counts.
The Guardian, covering the case, frames it as one of the first major U.S. criminal matters where AI-generated music is used as a tool inside a streaming-fraud operation, rather than as the subject of an authorship or copyright dispute. That framing helps clarify what’s at stake: AI isn’t just changing how songs are made; it’s changing the economics of misconduct.
Two features of AI-generated catalogs matter in the government’s telling:
- Volume: a massive number of tracks increases the surface area of the fraud.
- Variety: a large, varied catalog can avoid the obvious red flags of looping a small set of songs.
A single track that receives an implausible number of plays is easy to flag. A large catalog each receiving a plausible trickle is harder—especially when the long tail of legitimate streaming already contains countless niche tracks with modest play counts.
The Guardian, covering the case, frames it as one of the first major U.S. criminal matters where AI-generated music is used as a tool inside a streaming-fraud operation, rather than as the subject of an authorship or copyright dispute. That framing helps clarify what’s at stake: AI isn’t just changing how songs are made; it’s changing the economics of misconduct.
The real innovation: synthetic supply + synthetic demand
The alleged scheme pairs two forms of automation:
- Synthetic supply: AI-generated tracks uploaded at scale.
- Synthetic demand: bot accounts generating streams at scale.
Either one alone is limited. A huge catalog without plays earns little. A bot network without tracks is useless. Put them together, and the operation can resemble a functioning label—except the “roster” is machine-made and the “fans” are scripted.
- Synthetic supply: AI-generated tracks uploaded at scale.
- Synthetic demand: bot accounts generating streams at scale.
Either one alone is limited. A huge catalog without plays earns little. A bot network without tracks is useless. Put them together, and the operation can resemble a functioning label—except the “roster” is machine-made and the “fans” are scripted.
Key Insight
Synthetic supply (AI tracks) plus synthetic demand (bots) can mimic the “long tail” of legitimate catalogs—making fraud harder to spot at a glance.
The timeline and the totals: what we know from the public record
Public reporting and DOJ materials describe activity spanning 2017–2024. That seven-year window matters because it suggests persistence—an operation with staying power, not a one-off stunt.
The other numbers are stark:
- “Hundreds of thousands” of AI-generated songs (DOJ description)
- “Billions” of streams generated by bots (DOJ description)
- More than $8 million in royalties “fraudulently obtain[ed]” (DOJ)
- $8,091,843.64 forfeiture amount in the plea agreement (DOJ)
- Up to 5 years maximum imprisonment under the plea (DOJ)
- Sentencing scheduled for July 2026 (DOJ; multiple reports cite July 29, 2026)
Taken together, these figures depict a scheme that treated streaming like a programmable financial instrument: generate content cheaply, route it through distribution, and simulate the consumer behavior that unlocks payment.
The other numbers are stark:
- “Hundreds of thousands” of AI-generated songs (DOJ description)
- “Billions” of streams generated by bots (DOJ description)
- More than $8 million in royalties “fraudulently obtain[ed]” (DOJ)
- $8,091,843.64 forfeiture amount in the plea agreement (DOJ)
- Up to 5 years maximum imprisonment under the plea (DOJ)
- Sentencing scheduled for July 2026 (DOJ; multiple reports cite July 29, 2026)
Taken together, these figures depict a scheme that treated streaming like a programmable financial instrument: generate content cheaply, route it through distribution, and simulate the consumer behavior that unlocks payment.
Billions
DOJ description of the scale of bot-driven streams used to manufacture listening activity.
Up to 5 years
Maximum penalty cited in public DOJ coverage for the pleaded charge of conspiracy to commit wire fraud.
Why the number differences matter
Some readers fixate on why early coverage mentioned $10 million while the plea centers on $8 million and a forfeiture of $8,091,843.64. The answer is procedural rather than mysterious. Charging documents often reflect preliminary estimates and broader alleged conduct; pleas tend to settle on amounts tied to provable gains, negotiated stipulations, and the government’s forfeiture case. (Forbes; DOJ)
The key point remains: millions of dollars moved through a system designed to reward actual listening.
The key point remains: millions of dollars moved through a system designed to reward actual listening.
Who pays the price: artists, songwriters, and the credibility of discovery
The most immediate victims are the people the DOJ names: legitimate musicians and songwriters. In a pro‑rata pool, diversion is not abstract; it can be mechanical. Fraudulent streams take a share of the pot, leaving less for everyone else.
Yet the second-order harm may be broader: discovery itself becomes less trustworthy. Streaming platforms are not just payment pipes; they are recommendation engines. When automated activity inflates engagement signals, it can distort what gets surfaced.
Even readers who don’t care about royalty formulas should care about whether “popular” means popular. If bots can manufacture “billions” of plays, then play counts become less like audience measurement and more like a manipulable metric—closer to a billboard you can rent than a crowd you can earn.
Yet the second-order harm may be broader: discovery itself becomes less trustworthy. Streaming platforms are not just payment pipes; they are recommendation engines. When automated activity inflates engagement signals, it can distort what gets surfaced.
Even readers who don’t care about royalty formulas should care about whether “popular” means popular. If bots can manufacture “billions” of plays, then play counts become less like audience measurement and more like a manipulable metric—closer to a billboard you can rent than a crowd you can earn.
Multiple perspectives: why platforms, too, are in a bind
Platforms have incentives to fight fraud, but they also have reasons to be cautious and quiet. Publicly identifying fraudulent catalogs can invite legal fights with distributors and aggregators. Over-enforcement can also punish legitimate artists whose listening patterns look unusual—say, a niche genre with intense repeat listening.
That tension helps explain why a criminal case matters. Enforcement through private moderation has limits. A DOJ action signals a willingness to treat streaming fraud not as a platform policy violation, but as a crime with real penalties.
That tension helps explain why a criminal case matters. Enforcement through private moderation has limits. A DOJ action signals a willingness to treat streaming fraud not as a platform policy violation, but as a crime with real penalties.
Editor's Note
This article’s central claim about pay dilution follows the logic of pro‑rata pools: a bigger stream-count denominator can shrink legitimate shares when fake streams don’t add meaningful revenue.
Detection and enforcement: what this case signals for the industry
A guilty plea does not automatically reveal how the operation was detected, what internal platform systems flagged it, or what evidence formed the backbone of the government’s case. The DOJ’s press release focuses on the outline of conduct and the harm. Still, the plea itself functions as a warning flare.
First, it suggests federal authorities are prepared to pursue streaming fraud with the same seriousness applied to other digitally mediated financial crimes. The pleaded charge—conspiracy to commit wire fraud—places the conduct in a familiar legal category: deception using electronic communications to obtain money.
Second, it suggests that “AI-generated” will not serve as a fog machine for accountability. The government’s language treats AI as an enabler, not an excuse. That framing matters for future cases. As automation becomes more common in content creation, the line between legitimate experimentation and fraudulent monetization will hinge on intent and deception, not the novelty of the tool.
First, it suggests federal authorities are prepared to pursue streaming fraud with the same seriousness applied to other digitally mediated financial crimes. The pleaded charge—conspiracy to commit wire fraud—places the conduct in a familiar legal category: deception using electronic communications to obtain money.
Second, it suggests that “AI-generated” will not serve as a fog machine for accountability. The government’s language treats AI as an enabler, not an excuse. That framing matters for future cases. As automation becomes more common in content creation, the line between legitimate experimentation and fraudulent monetization will hinge on intent and deception, not the novelty of the tool.
Practical implications for creators and rights-holders
No individual artist can police the denominator in a pro‑rata system. Still, the case underscores a few practical realities:
- Transparency pressure will grow. Creators and publishers may demand clearer reporting on fraud adjustments and payout integrity.
- Distributor due diligence will matter more. When “hundreds of thousands” of tracks can appear quickly, the burden shifts toward vetting and monitoring at ingestion points.
- Metrics will be treated with more skepticism. Play counts and follower numbers may lose value as proof of organic demand—especially for unknown catalogs with unusual patterns.
- Transparency pressure will grow. Creators and publishers may demand clearer reporting on fraud adjustments and payout integrity.
- Distributor due diligence will matter more. When “hundreds of thousands” of tracks can appear quickly, the burden shifts toward vetting and monitoring at ingestion points.
- Metrics will be treated with more skepticism. Play counts and follower numbers may lose value as proof of organic demand—especially for unknown catalogs with unusual patterns.
What creators and rights-holders may push for next
- ✓Clearer reporting on fraud adjustments and payout integrity
- ✓Stronger distributor/aggregator vetting at ingestion points
- ✓Greater skepticism about plays and follower counts as proof of organic demand
A case study in the economics of “infinite music”
Smith’s alleged operation is best understood as an early stress test of an “infinite music” economy: when supply approaches zero cost, the bottleneck shifts from creation to attention—and attention can be simulated.
That doesn’t mean AI music is inherently suspect. Many artists use AI tools as part of composition, sound design, or ideation without trying to defraud anyone. The case is a reminder that the same reduction in friction that empowers creators also lowers the barrier for abuse.
The pro‑rata system amplifies the problem because it treats all counted streams as competing claims on the same pool. In that sense, the scheme described by the DOJ resembles a kind of private tax on legitimate listening—an extraction of value made possible by scale.
What happens next will matter. Sentencing in July 2026 will answer one question—how costly this conduct is for one defendant. The larger question belongs to the industry: whether streaming’s payment architecture and measurement culture can withstand a world where content and consumption can both be fabricated cheaply.
That doesn’t mean AI music is inherently suspect. Many artists use AI tools as part of composition, sound design, or ideation without trying to defraud anyone. The case is a reminder that the same reduction in friction that empowers creators also lowers the barrier for abuse.
The pro‑rata system amplifies the problem because it treats all counted streams as competing claims on the same pool. In that sense, the scheme described by the DOJ resembles a kind of private tax on legitimate listening—an extraction of value made possible by scale.
What happens next will matter. Sentencing in July 2026 will answer one question—how costly this conduct is for one defendant. The larger question belongs to the industry: whether streaming’s payment architecture and measurement culture can withstand a world where content and consumption can both be fabricated cheaply.
Conclusion: the fight is over trust, not technology
The Michael Smith plea is a clean narrative in a messy domain: a man, a method, a number, and a clear federal charge. Hundreds of thousands of AI-generated songs. Billions of bot-driven streams. More than $8 million diverted, with $8,091,843.64 on the forfeiture line. (DOJ)
The deeper story is what those numbers reveal about streaming itself. Under pro‑rata payouts, fraud doesn’t stay contained. It leaks into the economics of every artist who depends on the same pool and into the credibility of the metrics listeners use to decide what is worth hearing.
AI will keep making music cheaper to produce. The industry can live with that. The harder challenge is making truth—real listening, real demand, real cultural signal—expensive enough to fake that fraudsters stop trying.
The deeper story is what those numbers reveal about streaming itself. Under pro‑rata payouts, fraud doesn’t stay contained. It leaks into the economics of every artist who depends on the same pool and into the credibility of the metrics listeners use to decide what is worth hearing.
AI will keep making music cheaper to produce. The industry can live with that. The harder challenge is making truth—real listening, real demand, real cultural signal—expensive enough to fake that fraudsters stop trying.
1) Who is Michael Smith, and what did he plead guilty to?
Michael Smith is a man from Cornelius, North Carolina. According to the U.S. Department of Justice, he pleaded guilty to conspiracy to commit wire fraud. Prosecutors say the conspiracy involved distributing large volumes of AI-generated songs and using automated bot accounts to stream them at massive scale to generate royalty payments.
2) How much money did prosecutors say the scheme made?
The DOJ stated Smith “fraudulently obtain[ed] more than $8 million in royalties. The plea agreement also includes a forfeiture amount of $8,091,843.64. Earlier reporting around the September 2024 charge referenced “more than $10 million,” reflecting how estimates can change between initial charges and plea-stage stipulations.
3) How can fake streams reduce what real artists get paid?
Many platforms primarily use pro‑rata payouts: a pool of net revenue is divided based on share of total streams. If fraudulent activity adds streams without adding real revenue, the total stream count (the denominator) rises and the fraudster takes a share of the pool—meaning less is available for legitimate artists and songwriters whose work was actually streamed by real listeners.
4) Why does AI matter in this case specifically?
The DOJ’s description suggests AI functioned as an enabler of scale. Generating “hundreds of thousands” of tracks by traditional means would be slow and costly. AI can produce large catalogs quickly, which—combined with bots—can make a fraud operation harder to spot because it resembles the normal long tail of streaming catalogs.
5) What time period did the alleged activity cover?
DOJ-aligned reporting describes the conduct as spanning 2017 through 2024. That matters because it indicates a long-running operation rather than a short burst, potentially affecting multiple payout periods across many years of streaming revenue distribution.
6) When is sentencing, and what punishment is possible?
The DOJ indicates sentencing is scheduled for July 2026 (several reports cite July 29, 2026). Under the plea to conspiracy to commit wire fraud, public DOJ coverage cites a maximum penalty of up to five years in prison, though actual sentences depend on federal guidelines and judicial discretion.
7) Is this mainly a copyright case about AI music?
No. The public DOJ materials frame it as a fraud case, not an authorship or copyright dispute. The central allegation is deception—using bots and fabricated listening to obtain money. AI-generated music is described as part of the operational toolkit that helped produce a massive catalog, rather than the legal issue at the heart of the charge.
T
About the Author
TheMurrow Editorial is a writer for TheMurrow covering entertainment.
Frequently Asked Questions
Who is Michael Smith, and what did he plead guilty to?
Michael Smith is a man from Cornelius, North Carolina. According to the U.S. Department of Justice, he pleaded guilty to conspiracy to commit wire fraud. Prosecutors say the conspiracy involved distributing large volumes of AI-generated songs and using automated bot accounts to stream them at massive scale to generate royalty payments.
How much money did prosecutors say the scheme made?
The DOJ stated Smith “fraudulently obtain[ed] more than $8 million in royalties. The plea agreement also includes a forfeiture amount of $8,091,843.64. Earlier reporting around the September 2024 charge referenced “more than $10 million,” reflecting how estimates can change between initial charges and plea-stage stipulations.
How can fake streams reduce what real artists get paid?
Many platforms primarily use pro‑rata payouts: a pool of net revenue is divided based on share of total streams. If fraudulent activity adds streams without adding real revenue, the total stream count (the denominator) rises and the fraudster takes a share of the pool—meaning less is available for legitimate artists and songwriters whose work was actually streamed by real listeners.
Why does AI matter in this case specifically?
The DOJ’s description suggests AI functioned as an enabler of scale. Generating “hundreds of thousands” of tracks by traditional means would be slow and costly. AI can produce large catalogs quickly, which—combined with bots—can make a fraud operation harder to spot because it resembles the normal long tail of streaming catalogs.
When is sentencing, and what punishment is possible?
The DOJ indicates sentencing is scheduled for July 2026 (several reports cite July 29, 2026). Under the plea to conspiracy to commit wire fraud, public DOJ coverage cites a maximum penalty of up to five years in prison, though actual sentences depend on federal guidelines and judicial discretion.
Is this mainly a copyright case about AI music?
No. The public DOJ materials frame it as a fraud case, not an authorship or copyright dispute. The central allegation is deception—using bots and fabricated listening to obtain money. AI-generated music is described as part of the operational toolkit that helped produce a massive catalog, rather than the legal issue at the heart of the charge.
