SIAgentNews.com
← All news
DevelopingAgentsSecurity

Coinbase AI blocks $4.7 million in Elon Musk deepfake crypto scam

An autonomous AI agent intercepted a massive fraud attempt where scammers used a fake Elon Musk video to trick 500,000 viewers into sending crypto.

· First reported by Cointribune

Cinematic film still: picture for the story 'Coinbase AI blocks $4.7 million in Elon Musk deepfake crypto scam'
AI-made cinematic film still for this story. No real people or products are shown.
91%of at-risk funds blocked by AI

The short version

  • Scammers used a deepfake of Elon Musk to promise doubled crypto deposits to over 500,000 YouTube viewers.
  • Coinbase AI blocked $4.7 million in suspicious transfers without any human intervention.
  • The scam involved no technical breaches; victims voluntarily authorized the transactions based on the fake broadcast.
  • The FBI recently warned about ARTEX, an autonomous AI tool capable of conducting cyberattacks in under three hours.

Why it matters

This event demonstrates that traditional fraud filters are no longer enough to stop sophisticated, high-speed social engineering. It shows that defensive AI is becoming a necessary shield against increasingly automated and convincing cyberattacks.

The deepfake heist

Cybercriminals recently orchestrated a sophisticated scam by hijacking the narrative of a Tesla 'flying car' demonstration. They launched a fake live stream on YouTube that perfectly mimicked an official broadcast, featuring a deepfake of Elon Musk. This high-quality impersonation was designed to deceive a massive audience of more than 500,000 people.

The scam functioned by promising viewers that they could instantly double their cryptocurrency deposits. To facilitate this, the scammers displayed specific crypto wallet addresses and a QR code directly on the screen. The goal was to create a sense of urgency, prompting a rush of people to send their assets to the fraudulent addresses.

Within a very short window of just two hours, hundreds of Coinbase customers attempted to move millions of dollars to the scammers. Because the victims were convinced they were seizing a legitimate financial opportunity, they were not being tricked by a hack. Instead, they were voluntarily authenticated and approved the transactions themselves, moving assets like BTC and ETH.

Autonomous AI defense

Traditional anti-fraud filters often struggle to distinguish between these types of voluntary transfers and legitimate user activity. Because there was no technical breach, no passwords were stolen, and no accounts were hacked, standard security measures were less effective. Coinbase responded by deploying an AI-driven scam prevention agent to handle the surge of suspicious activity.

This specific AI agent operated entirely without any human intervention, acting as an autonomous shield. It was designed to identify and intercept high-risk transfers in real time as they were being initiated. By operating autonomously, the system could react at the speed of the scam, which was necessary to protect the users' funds before they left the platform.

The technology successfully managed to block about 91% of the at-risk funds. This resulted in the preservation of $4.7 million that would have otherwise been lost to the scammers. As Cointribune reported, the system's ability to operate without human oversight marks a significant escalation in how platforms defend against AI-driven fraud.

The rise of ARTEX

The use of an Elon Musk deepfake is part of a broader trend where cybercriminals are using advanced artificial intelligence to conduct attacks. The FBI recently issued a warning regarding the malicious exploitation of a tool called ARTEX. This is an autonomous penetration testing platform that uses large language models and AI agents to find vulnerabilities.

While ARTEX was originally designed for security testing, state-supported actors and criminals are now hijacking the open-source tool. They use it to automate the entire process of a cyberattack, which can sometimes be completed in under three hours. This speed makes it incredibly difficult for traditional security teams to react before the damage is done.

The FBI noted that ARTEX has already been used in several high-impact incidents. It was involved in attacks against South Korean financial institutions, which exposed the data of 67,000 people. Additionally, it was used to compromise the ConnectWise supply chain, affecting more than 2,250 Microsoft 365 client environments. These examples show how quickly AI can scale cybercrime.

A new security war

The transition from manual scams to autonomous AI attacks represents a major shift in the global cybercrime landscape. We are moving into an era where the 'crypto security war' is being fought between different types of artificial intelligence. On one side, criminals use tools like ARTEX to automate intrusions; on the other, platforms use agents like Coinbase's to provide real-time protection.

This incident proves that defensive AI can effectively neutralize multi-million-dollar scams by breaking the psychological grip of the fraud. However, the complexity of these attacks means that technical defenses alone are not a complete solution. The human element remains a significant factor, as the victims in this case were manipulated by the perceived authority of a famous figure.

Moving forward, the success of these autonomous shields will likely determine how much trust users can place in digital assets. As scams become more convincing and faster, the ability of a platform to intervene automatically becomes a primary line of defense. For now, the most effective barrier remains investor vigilance against any promises of easy financial gains.

What to watch

  • How other financial institutions will adopt autonomous AI shields to counter ARTEX-style attacks.
  • Whether more platforms will release data on the success rate of AI-driven fraud prevention.

Try it

Check your crypto wallet for any unauthorized 'test' transactions or unusual activity from unknown addresses.

Sources and links

First reported by Cointribune. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news