OpenAI and Paradigm Release EVMbench to Explore AI Agent Applications in Smart Contract Security

BlockBeats News, February 19, OpenAI and crypto investment firm Paradigm announced a partnership to launch EVMbench, designed to assess an AI Agent's capabilities in the smart contract security space, including identifying, patching, and exploiting critical vulnerabilities.

The two parties stated that as AI's ability in code generation and execution improves, it could be leveraged by attackers and also serve as a defense tool. Therefore, it is necessary to test AI systems' security capabilities in an environment that has real economic implications. EVMbench will provide various testing modes, including vulnerability detection, contract mutation, exploit eradication, and simulating full attack flows in a sandboxed chain environment.

At the tool's launch, DeFi protocol Moonwell and cross-chain liquidity protocol CrossCurve recently suffered smart contract vulnerability attacks, sparking concerns about AI-assisted code security. EVMbench is said to be built on 120 curated vulnerability samples, covering multiple public audit and competition cases. (The Block)