OpenAI has launched EVMbench, a new framework developed with Paradigm, to evaluate whether artificial intelligence can effectively secure smart contracts on blockchains like Ethereum. The tool assesses AI's ability to identify, exploit, and fix vulnerabilities in these self-executing codes. This initiative aims to set standards for AI in blockchain security amid growing stakes in decentralized finance.
OpenAI is expanding its involvement in cryptocurrency security through the introduction of EVMbench, a benchmark designed to gauge the effectiveness of modern AI systems in handling smart contract vulnerabilities. Smart contracts, which are self-executing programs on blockchains such as Ethereum, power decentralized exchanges, lending protocols, and various onchain financial services. Once deployed, these contracts are generally immutable, making any security flaws potentially severe.
Collaborating with crypto investment firm Paradigm, OpenAI created EVMbench using real-world vulnerabilities identified in past audits and security competitions. The framework evaluates AI performance in three key areas: detecting security bugs, exploiting them in a controlled setting, and repairing the code without disrupting functionality.
The benchmark seeks to provide a standardized method for assessing AI's role in blockchain security, particularly as decentralized finance manages billions in user funds. OpenAI emphasized the importance of this work in a blog post, stating: “Smart contracts routinely secure $100B+ in open-source crypto assets. As AI agents improve at reading, writing, and executing code, it becomes increasingly important to measure their capabilities in economically meaningful environments, and to encourage the use of AI systems defensively to audit and strengthen deployed contracts.”
This launch highlights ongoing efforts to integrate AI defensively into crypto ecosystems, where the value at risk continues to grow.