Coinbase disclosed on November 17, 2026, that its AI-powered scam prevention system blocked approximately $4.7 million in attempted cryptocurrency transfers connected to a deepfake livestream scam. Fraudsters used AI-generated footage to impersonate a celebrity during a fake Tesla event and promised to return twice the amount of crypto sent to designated wallets. Coinbase said its system prevented around 91% of the attempted transfers. The incident highlights how crypto fraud prevention is expanding beyond account-compromise detection to identify transactions that customers authorize themselves after being deceived.
How did scammers use deepfakes to persuade users to transfer funds?
According to Coinbase, scammers created a YouTube livestream that appeared to be associated with an official Tesla event. The stream used AI-generated footage of a well-known business figure and displayed wallet addresses and QR codes. Viewers were told that sending cryptocurrency to the specified addresses would result in receiving twice the amount.
The livestream attracted more than 500,000 viewers. Within approximately two hours, hundreds of Coinbase customers attempted to send funds to the scam addresses. The scheme combined celebrity impersonation, a time-sensitive event and unrealistic financial promises to pressure viewers into authorizing irreversible transfers.
How did Coinbase's AI system identify risky transactions?
Coinbase said its AI defenses scored outgoing transfers for risk patterns associated with fake giveaways, brand impersonation and newly created wallet addresses. The system also deployed a dedicated scam-prevention agent to examine transaction context and explain how the suspected fraud worked.
The company reported that its system blocked around 91% of attempted transfers, representing approximately $4.7 million. This figure refers to attempted transfers that were prevented, not confirmed losses of the same amount. AI systems also cannot guarantee that every scam will be detected, particularly when customers continue with a transfer after receiving a warning.
How should crypto platforms combine AI and KYT to prevent scams?
Exchanges can combine device-risk signals, account behavior, recipient-address reputation and blockchain transaction analysis to identify suspicious payments. Patterns such as multiple users sending funds to the same new address, sudden aggregation of small transfers or rapid movement to high-risk services may require additional review.
AI can help identify changing scam narratives and behavioral patterns, while KYT provides wallet associations, fund-flow information and historical risk indicators. Together, these capabilities can improve detection before funds leave a platform. However, firms should retain human review, customer appeal and false-positive handling procedures. The incident shows that fraud prevention must address both stolen accounts and customers who are manipulated into authorizing payments themselves.