Coinbase says it blocked $4.7 million in Musk deepfake scam transfers
According to BleepingComputer, Coinbase says it blocked $4.7 million in customer transfers to scammers promoting a fake giveaway built on an AI deepfake of Elon Musk.
On this page
According to BleepingComputer, Coinbase says it blocked $4.7 million in cryptocurrency transfers to scammers promoting a fake giveaway that used an artificial-intelligence (AI) deepfake of Elon Musk.
What happened
The report says the deepfake ran during Tesla’s “flying car” demonstration, when scammers created a YouTube livestream that appeared to broadcast the event. That stream attracted more than 500,000 viewers. Hundreds of Coinbase customers attempted to send millions of dollars to the scammers over about two hours, and Coinbase’s AI-powered fraud-prevention systems blocked about 91% of the attempted transfers, according to the report.
Why it matters
The report says victims voluntarily made the crypto transfer after believing the livestream was real, rather than through compromised accounts or stolen credentials. Coinbase’s scam prevention system uses machine learning models to analyze outgoing cryptocurrency transfers in real time. When a suspicious transfer is detected, an AI agent engages the customer in a short conversation explaining the characteristics of the suspected scam.
How much ran without a person
Coinbase claims the entire detection and intervention process operated automatically, without human involvement. The report says the AI agent had never encountered this particular campaign before but recognized the underlying scam techniques.
What to do
Treat the doubling promise as the signal: the fraudulent stream told viewers that any cryptocurrency they sent would be doubled. In this campaign the customers sent the funds themselves after believing the livestream was real, so the moment that matters is the transfer itself, where Coinbase says an AI agent opens a short conversation about the suspected scam.
Sources
Related reading
3 postsDiscussion
Spotted an error? Send corrections to [email protected] .