The post OpenAI and Paradigm Launch EVMbench for Improving Smart Contract Security appeared on BitcoinEthereumNews.com.

OpenAI and Paradigm partnered to launch EVMbench. The EVMbench protocol will detect and fix Smart Contract vulnerabilities. OpenAI aims to set a measurable standard for assessing AIAgents in blockchain security. American intelligence research organization OpenAI, in collaboration with Paradigm, a crypto investment firm, has launched EVMbench, a benchmarking tool to evaluate AI systems’ ability to identify, exploit, and repair vulnerabilities in Smart Contracts deployed on Ethereum-compatible blockchains. Introducing EVMbench—a new benchmark that measures how well AI agents can detect, exploit, and patch high-severity smart contract vulnerabilities. https://t.co/op5zufgAGH — OpenAI (@OpenAI) February 18, 2026 A Research-based Product According to a paper published by OpenAI, the AI company noted that EVMbench draws on 120 curated vulnerabilities. Vulnerabilities are sourced from 40 security audits, including sponsored open-code audit competitions and the security review conducted by Tempo, a Layer-1 blockchain co-developed by Paradigm and Stripe.  Meanwhile, EVMbench’s design allows it to test AI agents across three core functions, including detecting security flaws, patching vulnerable code without breaking contract functionality, and executing controlled fund-draining attacks in a sandboxed blockchain environment. For context, Smart Contracts are self-executing programs that run on blockchains. They are typically deployed to power decentralized exchanges, lending platforms, and a wide range of on-chain financial applications, particularly in the DeFi ecosystem. It is worth noting that Smart Contracts are typically immutable after deployment. Therefore, a ‘minor’ vulnerability can have significant implications, including substantial financial losses for users and protocols. The EVMbench has Multiple Capabilities Notably, EVMbench’s launch follows the recent exploit of the Moonwell DeFi protocol, involving vulnerable code written with AI assistance, and the CrossCurve cross-chain liquidity protocol, which was recently compromised through a smart contract vulnerability. Both incidents resulted in approximately $3 million in losses across multiple networks. According to OpenAI, the new product will include multiple capability modes covering…

---

Filed under: News - @ February 20, 2026 6:22 pm