Apple Music flagged 2 billion fraudulent streams in 2025 while AI-powered bot farms have already stolen over $10M in royalties

Apple Music identified and demonetized approximately 2 billion fraudulent streams throughout 2025, representing roughly $17 million in royalties that would have been improperly distributed. In the most prominent criminal case, Michael Smith of North Carolina pleaded guilty to using AI-generated songs and bot accounts to extract over $10 million in fraudulent royalty payments from Spotify, Apple Music, and Amazon Music. Meanwhile, Deezer found that up to 85% of streams generated by fully AI-produced music were flagged as fraudulent in 2025. The fraud infrastructure has industrialized: bot farms use hundreds of physical smartphones programmed to stream 24/7, with AI systems rotating through VPNs and proxies to simulate geographically diverse human listening patterns. Every dollar paid to a fraudulent stream is a dollar taken from a legitimate artist. Under the pro-rata model used by major platforms, royalties come from a single pool — so when bot farms inflate stream counts for fake tracks, they are not generating new money but redistributing existing money away from real musicians. The $17 million Apple Music demonetized represents royalties that would have otherwise been split among artists whose fans were actually listening. For an independent artist earning $3,000-$5,000 annually from streaming, even a 1-2% reduction in their share of the royalty pool — caused by fraud dilution — represents meaningful lost income. The problem is worse than the headline numbers suggest because detected fraud is only the fraud that gets caught. Sophisticated operations specifically design their bot behavior to stay below detection thresholds, streaming each fake track just enough times to avoid triggering anomaly alerts. The structural reason this persists is that fraud detection is an arms race where the fraudsters have asymmetric advantages. Generating a fake song with AI costs effectively nothing. Uploading it through a distributor requires minimal verification. Running a bot farm requires modest technical sophistication that is increasingly commoditized. Meanwhile, platforms must process billions of streams in real time, balance false positive rates (flagging legitimate niche artists as bots) against false negative rates (missing sophisticated fraud), and do so while maintaining user experience. The financial incentive structure also works against aggressive enforcement: platforms report total stream counts to investors and advertisers, and aggressive purging of fraudulent streams would reduce those headline numbers. The Michael Smith case was notable precisely because it was the first U.S. criminal prosecution — indicating that law enforcement is only beginning to treat streaming fraud as a serious crime.