Back Aiweekly.Co Coinbase AI blocks $4.7M in Musk deepfake 'flying car' scam
Coinbase's AI flagged the deepfake giveaway scheme with no prior training on that attack type, using signals including brand impersonation patterns and newly created destination wallets.
No accounts were compromised: every transfer was customer-initiated and voluntarily confirmed, bypassing all traditional unauthorized-access fraud controls.
The deepfake livestream drew over 500,000 concurrent viewers during a live Tesla event, exploiting a high-attention moment to accelerate victim conversion at scale.
Coinbase says its fraud systems blocked $4.7 million in cryptocurrency transfers to a two-hour YouTube livestream that featured an AI-generated Elon Musk promoting a Tesla 'flying car' and offering to double any crypto sent to the wallet addresses and QR codes on screen.
The broadcast pulled in more than 500,000 viewers. Hundreds of Coinbase customers tried to send funds. The exchange, according to BleepingComputer , stopped roughly 91% of the attempted transfers before they left customer accounts.
What makes the case unusual is that nothing the transactions was unauthorized. 'Most fraud defenses are built to spot activity the customer didn't authorize. That's not what happened here,' Coinbase said. 'The customer was logged in, the customer pressed confirm, and the customer believed it was the opportunity of a lifetime.'
The company's machine-learning model flags outbound transfers that match the shape of giveaway and impersonation scams, then hands the session to an AI agent that talks the customer through what it sees. Coinbase says the system had never encountered this specific campaign. '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 did not break out how much of the that still went through represents real customer losses. It is one of more than two dozen deepfake stories we have tracked in the last 90 days , and the pattern keeps tilting the same way: the fake video is the bait, but the loss lands in a self-authorized transfer.
What others are reporting
Coverage cluster as of 2h after publish
Unchained Crypto Read → Details how the AI intervened via in-app chat explaining the scam before transfer completion, and that the model flagged the scheme despite having no prior deepfake giveaway training data. Nobody was hacked. Customers logged in and approved the transfers themselves, which makes this kind of fraud hard to catch.
Details how the AI intervened via in-app chat explaining the scam before transfer completion, and that the model flagged the scheme despite having no prior deepfake giveaway training data.
Nobody was hacked. Customers logged in and approved the transfers themselves, which makes this kind of fraud hard to catch.
Cryptopolitan Read → Frames this as a six-year evolution from Coinbase's 2020 Twitter-hack response, contrasting blacklisting after funds moved with proactive pre-transfer interception now. Our AI-powered scam defenses stopped $4.7 million from leaving customers' accounts.
Frames this as a six-year evolution from Coinbase's 2020 Twitter-hack response, contrasting blacklisting after funds moved with proactive pre-transfer interception now.
Our AI-powered scam defenses stopped $4.7 million from leaving customers' accounts.
FinanceFeeds Read → Adds industry-scale context: crypto fraud hit $11.4B in 2025 total losses, with AI-related financial crimes alone accounting for $893M that year. Coinbase addressed the threat through an automated system combining machine-learning-based transaction monitoring with a dedicated scam prevention agent.
Adds industry-scale context: crypto fraud hit $11.4B in 2025 total losses, with AI-related financial crimes alone accounting for $893M that year.
Coinbase addressed the threat through an automated system combining machine-learning-based transaction monitoring with a dedicated scam prevention agent.
Bitcoin.com News Read → Emphasizes the two-hour attack window and machine-speed response, framing AI-on-AI fraud as the defining threat pattern for crypto exchanges going forward. In two hours, hundreds of Coinbase customers tried to send millions to the scammers. Our AI-powered scam defenses stopped $4.7 million from leaving customers' accounts.
Emphasizes the two-hour attack window and machine-speed response, framing AI-on-AI fraud as the defining threat pattern for crypto exchanges going forward.
In two hours, hundreds of Coinbase customers tried to send millions to the scammers. Our AI-powered scam defenses stopped $4.7 million from leaving customers' accounts.
Originally reported by bleepingcomputer.com
Original headline: Coinbase AI Blocks $4.7M in Transfers to AI-Generated Elon Musk 'Tesla Giveaway' Livestream Scam
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
