Coinbase revealed on November 17, 2026, that its artificial intelligence-powered fraud prevention system blocked approximately $4.7 million in customer cryptocurrency transfers tied to a deepfake Elon Musk giveaway scam.
The fraudsters launched a deceptive YouTube livestream during Tesla's flying car demonstration in October. It featured an AI-generated version of Elon Musk promoting a fake crypto giveaway that promised to double participants' deposits. The broadcast displayed wallet addresses and a QR code, directing viewers to transfer digital assets to receive rewards. More than 500,000 viewers watched the fake stream, and within roughly two hours, hundreds of Coinbase customers attempted to send millions of dollars to fraudulent addresses.
Because customers voluntarily authorized the transactions, traditional account security measures were insufficient. Coinbase instead deployed machine-learning models that examined transaction patterns, wallet characteristics, newly created addresses, impersonated brands, and promises of unusually high returns. A dedicated AI scam-prevention agent then engaged with customers to explain warning signs and assess the legitimacy of the giveaway. The system had not encountered this specific deepfake campaign during training, but it recognized familiar fraud indicators.
Coinbase said its automated defenses prevented approximately 91% of the attempted transfers, suggesting roughly $5.16 million may have been targeted in total. The company confirmed the incident did not involve stolen passwords, compromised credentials, or breaches of its security infrastructure.
The case highlights the growing threat of AI-generated impersonation scams. According to the FBI's 2025 Internet Crime Report, cryptocurrency-related fraud produced about $11.4 billion in reported losses, a 22% increase from 2024. Coinbase said it plans to increase investment in AI-powered fraud prevention as deepfake technology becomes more accessible.