Newsroom

Coinbase AI Blocks $4.7 Million in Transfers to Deepfake Musk Crypto Scam

18 November, 2026   /   News   /  AI   /   Tags:  deepfake, giveaway, livestream, scam, coinbase

Coinbase AI Blocks $4.7 Million in Transfers to Deepfake Musk Crypto Scam

Exchange systems stopped most customer attempts during a fake Tesla livestream giveaway that drew over half a million viewers

Coinbase reported that its artificial intelligence defenses prevented $4.7 million in cryptocurrency transfers from reaching scammers who promoted a fraudulent giveaway using a deepfake of Elon Musk. The incident unfolded around Tesla’s flying car demonstration, when fraudsters launched a YouTube livestream designed to appear as an official broadcast.

The stream featured an AI-generated version of Musk alongside on-screen wallet addresses and a QR code. Viewers were told any cryptocurrency sent to those addresses would be doubled. More than 500,000 people watched the livestream. Within roughly two hours, hundreds of Coinbase customers, including many long-time users, attempted to transfer millions of dollars to the addresses shown.

Customer-Authorized Transfers, Not Account Compromises

Coinbase stressed that no accounts were breached. No passwords were stolen and no credentials were phished. Customers logged in and approved the transfers themselves after being persuaded by the deepfake presentation and the promise of rapid returns. Traditional fraud systems built to catch unauthorized activity face greater difficulty with transactions that users deliberately authorize.

The exchange noted that its systems can identify risk signals in real time but cannot determine customer intent. Once a user is authenticated and confirms a transfer, distinguishing a genuine decision from deception based solely on the transaction data becomes challenging.

How the AI Detection and Intervention Worked

Machine-learning models scored outgoing transfers for risk indicators associated with giveaway schemes, brand impersonation, and newly created wallet addresses. When a transfer was flagged, an AI scam-prevention agent took over. The agent held short conversations with customers, named the specific scam pattern, explained the mechanics of the giveaway, and asked questions designed to reveal inconsistencies in the offer.

Each response from the agent carried a confidence score before reaching the customer. The system had not previously encountered this exact combination of a deepfake giveaway tied to a live product event. It nevertheless recognized the core elements: promises of multiplied deposits, limited-time urgency tied to a public event, and irreversible cryptocurrency transfers.

Coinbase stated that the automated process operated without human review of individual transactions. Approximately 91 percent of the attempted transfers linked to the scam were stopped before funds left customer accounts.

Rising Challenge of Deepfake Scams

The company observed that deepfake technology has lowered the cost and complexity of staging convincing frauds. What once required a production crew and significant preparation can now be launched quickly with a livestream slot and generative models. Scammers can relaunch similar schemes under new identities soon after one is disrupted.

In response, Coinbase said it is expanding investment in AI-driven risk systems. These include models that improve through continuous research so detection adapts to new formats, agents that review risky sends across multiple scam types, and faster labeling of emerging cases so learning occurs within hours rather than weeks.

Nobody legitimate will double your crypto. An offer that requires you to send funds first to receive more back is a scam, no matter who appears to be making it.
Coinbase

The exchange advised users to avoid sending cryptocurrency to addresses or QR codes displayed in videos or livestreams, to verify information through official channels rather than on-screen links, and to slow down when a transfer is flagged. It also stated that Coinbase will never request crypto transfers to claim prizes or giveaways, promise to multiply funds sent by users, or pressure customers to act before a livestream ends.

The $4.7 million blocked represents funds that would otherwise have left customer accounts during the brief window of the scam. Coinbase continues to refine these automated protections as deepfake and event-driven fraud techniques evolve.

Associated cryptocurrencies
Disclaimer
This article was generated by AI using information from multiple industry sources. It has not been reviewed or verified by a human editor and may contain inaccuracies, omissions, or misinformation. Readers are encouraged to independently verify any information before making decisions based on its content.
This article is for informational purposes only and does not constitute financial, legal, or investment advice. Cryptocurrency and related investments involve substantial risk, and past performance does not guarantee future results.