Why AI-driven cryptocurrency exploits are more dangerous than ever

Anthropic researchers discovered that three popular AI agents can autonomously exploit vulnerabilities in smart contracts, generating an estimated $4.6 million in simulated stolen funds.

They also discovered new vulnerabilities in recently published contracts on the blockchain, showing that AI-driven cyber attacks are now possible and profitable.

Sponsored

Sponsored

AI-driven cyberattacks have proven costly

In a blog post to publish On Monday, Anthropic revealed disturbing findings about artificial intelligence’s growing ability to target vulnerabilities. In smart contracts.

Their research revealed that three models AI-Cloud Opus 4.5, Sonet 4.5, and GPT-5-He was able to discover and exploit vulnerabilities in blockchain contracts. This resulted in $4.6 million in simulated stolen funds from contracts published after March 2025.

The AI ​​models also discovered two vulnerabilities in the recently launched contracts.

A vulnerability allowed attackers to manipulate a generic “Calculator” function, intended to determine token awards, to increase token balances. Another allowed the attackers to withdraw money Providing false beneficiary addresses.

GPT-5 was able to identify and exploit these issues at a cost of only $3,476. This number represents the cost of running an AI model to perform the attack in a simulated environment.

Sponsored

Sponsored

Since these exploits resulted in $4.6 million in stolen funds, the low costs required to implement them demonstrate that AI-driven cyberattacks are not only possible, but also cost-effective, making them profitable. It is attractive to potential cybercriminals.

Revenues from these AI-driven exploits are increasing at an alarming rate.

Exponential increase in exploitation profits

In the past year, the amount stolen by these attacks has almost doubled every 1.3 months.

This rapid growth shows how AI-driven exploits are quickly becoming more profitable and widespread. The models are improving in their ability to detect vulnerabilities and execute attacks more efficiently.

As the stolen money increases, it becomes more difficult for the organization to keep it. What is particularly worrying is that artificial intelligence can Now run it These attacks independently without human intervention.

It represents anthropic results A major change in cybersecurity. AI not only identifies vulnerabilities, but also designs and executes attack strategies autonomously with little oversight.

The implications go far beyond cryptocurrencies. Any software system with weak security is vulnerable to attack, starting with enterprise applications To financial services and beyond.