Incident sheet · /security

Filed
Read
4 min · 726 words
Severity
Routine

Coinbase blocks $4.7M in Musk deepfake giveaway transfers

Coinbase says its AI system blocked 91% of attempts worth $4.7 million in a fake Elon Musk giveaway during a Tesla livestream, but the final amount stolen remains unanswered.

Coinbase blocks $4.7 million in transfers to Elon Musk deepfake scam
FeedCoinbase blocks $4.7 million in transfers to Elon Musk deepfake scam — AI-generated

Event log

  1. Coinbase blocked about $4.7 million in attempted transfers during the roughly two-hour campaign.
  2. AI fraud controls stopped about 91% of attempts, according to Coinbase.
  3. Hundreds of customers attempted to send millions of dollars; many were longtime users.
  4. The fake stream ran during Tesla's “flying car” demonstration, watched by more than 500,000 viewers.
  5. After blocking 91%, Coinbase had not disclosed how many transfers completed or the final amount stolen.

Coinbase blocked $4.7 million in cryptocurrency transfers linked to an AI-generated Elon Musk giveaway, according to the exchange.

Coinbase said its fraud-prevention systems stopped about 91% of attempted transfers over roughly two hours. Hundreds of customers tried to send millions of dollars to the scammers. Many were longtime Coinbase users.

How did the fake giveaway work?

Scammers circulated a fraudulent YouTube livestream during Tesla's “flying car” demonstration. The counterfeit feed appeared to broadcast the event. Coinbase said the event attracted more than 500,000 viewers.

The stream displayed an AI-generated video of Musk, cryptocurrency wallet addresses, and a QR code. It promised viewers that any cryptocurrency sent would be doubled.

Customers then approved transfers themselves after mistaking the counterfeit broadcast for a genuine Musk promotion tied to the live product event seen during the demonstration.

Why did approved payments still trigger intervention?

Unlike traditional cryptocurrency theft involving compromised accounts or stolen credentials, this campaign relied on voluntary customer payments. Coinbase distinguished the mechanism clearly.

“Most fraud defenses are built to spot activity the customer didn't authorize,” Coinbase 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.”

The control challenge was unusually narrow: the account holder was logged in, consciously clicked confirmation, and still acted on a fabricated opportunity created by scammers.

What did Coinbase automate?

Coinbase's scam prevention system uses machine-learning models to analyze outgoing cryptocurrency transfers in real time. The models identify suspicious activity involving:

  • giveaway scams;
  • brand impersonation; and
  • newly created wallet addresses.

When a suspicious transfer appears, an AI agent opens a short conversation with the customer. It explains the suspected scam and asks questions intended to help recognition. The system assigns confidence scores to the agent's responses before presenting them to customers. Those scores guide decisions about whether additional safeguards are necessary.

This was an automated intervention rather than a manual review, according to Coinbase, which said human staff played no part in either detection or response. Coinbase's account did not identify the model architecture, version, false-positive rate, or number of customer interventions.

Could the agent handle a new campaign?

Coinbase said the agent had never encountered this particular campaign before. It nevertheless recognized three core techniques:

  • promises to multiply cryptocurrency;
  • time pressure tied to a live event; and
  • irreversible transfers.

“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.

The claim indicates pattern recognition rather than prior exposure to the exact fake livestream, while leaving unstated how broadly Coinbase's training data covered comparable scams.

How much reached the scammers?

BleepingComputer asked Coinbase two unanswered questions: how many transfers reached the scammers, and how much cryptocurrency they stole. The company had not responded by publication.

The $4.7 million figure measures blocked transfers, not a disclosed final loss. The 91% rate was approximate, and Coinbase gave no count of completed transfers.

Without those figures, the campaign's scale can be bounded at the platform's intervention level, but its final financial impact cannot be reconciled from disclosed information. That leaves Coinbase's core claim supported but unclosed: it stopped most attempts, while the amount successfully stolen remains undisclosed.

Why does the Musk lure persist?

Musk and his companies have long attracted cryptocurrency scammers. Their campaigns have included:

  • hijacking verified Twitter accounts;
  • sending fake giveaway invitations by email;
  • flooding TikTok with fraudulent promotions; and
  • using deepfake Musk videos to promote fraudulent investment platforms.

Another campaign involving compromised verified Twitter accounts stole more than $580,000 in cryptocurrency in one week.

Unlike the earlier $580,000 campaign, customers deliberately confirmed transfers after a convincing event rather than acting solely because verified accounts had been compromised by scammers.

Coinbase advises users to verify cryptocurrency promotions through official websites and social media accounts. It cautions against trusting livestreams or QR codes shown within videos.

For fraud teams, the lesson is direct: account authentication and customer confirmation do not by themselves establish that an apparently authorized transaction actually remains legitimate. No remediation cost or customer-recovery figure was disclosed. The automated response shows scale, but the missing loss total prevents a complete assessment of the campaign's financial cost.

via bleepingcomputer.com (Source)

Filed under

  • coinbase
  • cryptocurrency-fraud
  • deepfake-scams
  • fraud-prevention
  • ai-agents
Share this article:

More from Daniel Okafor

Daniel Okafor

Show full bio

Staff writer covering technology and product news in enterprise information technology and the leaders who run at CIO Monitor.

203 articles

The newsroom digest

One email a day with the stories that matter. No noise.

By subscribing you agree to the privacy policy.

Cross-references

‹ Previous articleNext article ›