OpenAI and Paradigm Release EVMbench to Enhance Smart Contract Security
- OpenAI and Paradigm launched EVMbench, a framework for evaluating AI agents in detecting, patching, and exploiting vulnerabilities in Ethereum smart contracts.
- The benchmark includes data from over 120 high-severity vulnerabilities sourced from real-world audits and competitive contests.
- Current AI models can exploit more than 70% of critical bugs, a significant improvement from less than 20% at the project’s inception.
- OpenAI expanded its Aardvark security research agent’s private beta and allocated $10M in API credits through its Cybersecurity Grant Program to bolster defensive crypto research.
EVMbench aims to enhance the security of smart contracts by leveraging AI capabilities to identify and mitigate vulnerabilities effectively. This initiative reflects OpenAI’s commitment to integrating AI with blockchain technology for improved cybersecurity.
The development of EVMbench highlights the potential for AI-driven solutions in securing decentralized finance platforms, addressing a critical need as over $5 billion has been lost due to exploits in recent years. Source
