OpenAI and Paradigm have launched EVMbench, a revolutionary platform designed to enhance decentralized finance (DeFi) security by testing advanced AI agents against one another. This initiative aims to address the critical vulnerabilities in smart contracts, particularly within Ethereum, where over $100 billion in crypto assets are at risk. Recent exploits, such as the $2.7 million loss at Moonwell Protocol, highlight the dangers of AI-generated vulnerabilities, necessitating robust defense mechanisms.
EVMbench simulates three roles: “Detect,” which identifies vulnerabilities, “Patch,” which aims to rectify them without introducing new flaws, and “Exploit,” where AI agents emulate attackers. Initial findings reveal AI’s strong detection capabilities but also emphasize that exploiting vulnerabilities is easier than fixing them. OpenAI’s private beta, Aardvark, complements EVMbench by autonomously monitoring and addressing vulnerabilities in real-time. This groundbreaking approach represents a pivotal step in safeguarding blockchain, ensuring that defenders outpace evolving threats in DeFi.
Source link
