Quick Take
- Coinbase's AI defenses blocked $4.7 million, or 91%, of attempted crypto transfers tied to an Elon Musk deepfake giveaway scam.
- The system stopped a customer's $100,000 payment, but the same user later bypassed its protection with a false explanation for a $5,000 transfer.
- The incident exposes a weakness in AI fraud prevention as exchanges struggle to protect customers who insist on sending money despite scam warnings.
Coinbase said its AI defenses blocked $4.7 million in attempted cryptocurrency transfers tied to an Elon Musk deepfake scam.
The crypto exchange said on Nov. 17 that its automated fraud-prevention system stopped about 91% of the money customers tried to send to scammers promoting a fake Tesla cryptocurrency giveaway.
The fraudsters had created a lookalike YouTube livestream during a Tesla demonstration, using an AI-generated impersonation of Musk to promise viewers twice the cryptocurrency they deposited into designated wallets.
More than 500,000 people watched the fraudulent broadcast, while hundreds of Coinbase customers attempted to transfer millions of dollars within roughly two hours.
Unlike conventional account compromises, the customers were willingly authorizing the transactions, believing they were participating in a legitimate promotion.
Coinbase said its AI system intercepted suspicious payments in real time, engaging customers before their funds could leave the platform without requiring human intervention.
Coinbase AI stops $100,000 transfer but misses customer's second attempt
Despite preventing most attempted transfers, Coinbase disclosed an incident that illustrates the difficulty of protecting customers determined to complete suspicious payments.
Will Megson, Coinbase's vice president of product working on risk, revealed that one customer initially attempted to transfer $100,000 to the scammers.
The exchange's AI agent identified the suspicious transaction, explained the fraudulent giveaway and convinced the customer to abandon the payment.
However, the customer subsequently returned and attempted another transfer, this time for $5,000.
One customer, two payment attempts
A Coinbase executive’s account of the AI agent’s intervention during the fake Tesla giveaway.
The warning interrupted one transfer. The customer came back, and the second payment went through.
This account describes one customer and does not establish who received the $5,000.
When questioned about the transaction, the customer falsely claimed the funds were intended for their daughter. The AI accepted the explanation and allowed the payment to proceed.
Megson acknowledged how difficult it is to distinguish legitimate payments from fraudulent transactions when customers provide misleading explanations.
The episode highlights a limitation of automated fraud detection: even when an exchange correctly identifies a scam, customers can still circumvent protections by changing their transaction details or misrepresenting their intentions.
Coinbase's system combines machine-learning models that identify suspicious transactions with an AI agent that examines the payment's context and questions customers about potential fraud.
The agent also assigns each response a confidence score before it reaches the customer.
The exchange said the system had not previously encountered the specific combination of a deepfake giveaway and a live product demonstration, but recognized familiar warning signs, including promises of guaranteed returns and irreversible payments.
Meanwhile, Coinbase said it is expanding its AI-powered fraud prevention efforts through improved detection models, automated reviews of suspicious transactions and faster identification of emerging scam techniques.


