Hundreds of cryptocurrency investors attempted to send money to scammers during a fraudulent Tesla livestream featuring an AI-generated impersonation of Elon Musk. Coinbase says its automated fraud detection system prevented approximately $4.7 million from reaching the criminals, blocking 91% of the funds customers tried to transfer.

 

 

The scheme unfolded during Tesla’s “flying car” demonstration in October 2026. Fraudsters created a deceptive YouTube broadcast that resembled the company’s official event, using deepfake footage of Musk to promote a cryptocurrency giveaway.

Viewers were instructed to send digital assets to wallet addresses displayed on the screen, with promises that their deposits would be returned at twice their original value. The broadcast also featured QR codes designed to simplify transfers.

According to Coinbase, the fraudulent livestream attracted more than 500,000 viewers. Within approximately two hours, hundreds of its customers attempted to transfer cryptocurrency to the scammers, including experienced users who had maintained accounts with the exchange for years.

Unlike incidents involving stolen passwords or compromised cryptocurrency wallets, this operation did not require attackers to access customer accounts. The victims initiated the transactions themselves after believing the promotional offer was genuine.

That distinction presented a challenge for conventional fraud detection systems, which frequently focus on identifying transactions initiated without an account holder’s authorization. In this case, the attempted payments came from authenticated customers who had approved the transfers.

Coinbase said its security infrastructure identified suspicious activity by analyzing outgoing transactions for indicators associated with fraudulent giveaways, impersonated brands, and newly created cryptocurrency addresses.

When the system detected a potentially fraudulent transfer, an AI-powered agent engaged the customer in a targeted conversation. Rather than issuing a standard warning, the agent explained the suspected deception and asked questions intended to help the customer recognize the scam.

The company also uses confidence scoring to evaluate the agent’s responses and determine whether additional safeguards are necessary.

Coinbase emphasized that its system had not previously encountered this particular campaign. Nevertheless, it recognized suspicious characteristics, including promises of unusually high returns, pressure to act before the livestream ended, and the irreversible nature of cryptocurrency transfers.

The exchange said the entire intervention operated automatically, without requiring human involvement. Its reported success rate and prevented losses are based on the company’s own assessment.

Coinbase has not disclosed how many transfers successfully reached the attackers or the total amount of cryptocurrency stolen. Consequently, while the company reports preventing $4.7 million in attempted payments, the financial losses associated with the fraudulent broadcast remain unknown.

Leave a Reply