During a Tesla "flying car" demonstration, scammers operated a counterfeit YouTube livestream that drew over 500,000 viewers. The fraudulent broadcast featured an AI-synthesized video of Musk alongside cryptocurrency wallet addresses and a QR code, claiming that any digital currency sent would be doubled. Coinbase reports it prevented $4.7 million in cryptocurrency transfers to the perpetrators.

image

Hundreds of Coinbase customers, including many established account holders, attempted to send millions of dollars to the scammers within approximately two hours. The exchange's AI-driven fraud-prevention systems detected the scheme and halted roughly 91% of the attempted transfers. This particular attack differs from conventional cryptocurrency theft because victims willingly initiated the transfers after being deceived by the livestream.

Coinbase noted that such scams present a unique detection challenge. "Most fraud defenses are built to spot activity the customer didn't authorize," the company explained. "That's not what happened here. The customer was logged in, the customer pressed confirm, and the customer believed it was the opportunity of a lifetime."

AI agents intervene to stop transfers

Coinbase's prevention system employs machine learning models to examine outgoing cryptocurrency transfers instantaneously, searching for red flags tied to giveaway schemes, impersonation tactics, and recently established wallet addresses. Upon identifying a suspicious transfer, an AI agent initiates a dialogue with the customer, describing the hallmarks of the suspected fraud and posing questions to help them recognize the deception.

The system generates confidence scores for the agent's responses before displaying them to users, permitting it to decide when heightened protections are warranted. Although the AI agent had not previously encountered this specific campaign, it recognized the underlying fraud patterns: promises to multiply cryptocurrency, urgency created by a live event, and irreversible transactions. Coinbase states the entire detection and intervention workflow operated mechanically, with no human oversight required.

"What's notable is that the agent had never seen this scam. Deepfake giveaways tied to a live product launch weren't in its training," Coinbase said.

Coinbase was contacted regarding the number of transfers that succeeded and the total amount of cryptocurrency lost, but has not yet provided a response. Musk and his enterprises have repeatedly served as targets for cryptocurrency fraud schemes, with bad actors seizing verified Twitter accounts, distributing fraudulent giveaway emails, and saturating TikTok with fake promotions. Deepfake videos of Musk have been weaponized to advertise illegitimate cryptocurrency investment schemes, and a separate incident involving compromised verified Twitter accounts resulted in over $580,000 in cryptocurrency losses within a single week.

To protect against such campaigns, users should confirm cryptocurrency promotions through legitimate websites and official social media channels rather than relying on livestreams or QR codes embedded in video content.