
EVMbench, developed by OpenAI and Paradigm, assesses AI agents’ detection, patching, and exploit simulation abilities for Ethereum smart contract security amid rising DeFi-related breaches.
OpenAI and crypto investment firm Paradigm have launched EVMbench, a benchmarking tool designed to evaluate AI agents’ performance in smart contract security tasks. The platform uses 120 curated vulnerability samples from past audits to test detection, patching, and exploit simulation capabilities within sandboxed environments. The initiative, unveiled shortly after the Moonwell and CrossCurve hack incidents, aims to provide a standardized framework for measuring AI effectiveness in identifying and fixing flaws in Ethereum Virtual Machine-compatible contracts.