OpenAI Launches EVMbench for Crypto Security Evaluation
- OpenAI introduced EVMbench, a system to assess AI agents’ ability to detect and fix security flaws in crypto tokens and smart contracts.
- The system was developed with Paradigm, focusing on vulnerabilities in code on Ethereum Virtual Machine-compatible blockchains.
- EVMbench evaluates AI performance in identifying weaknesses, demonstrating exploit methods, and applying fixes.
- OpenAI expanded its private beta of Aardvark and allocated $10 million in API credits for cybersecurity research through its grant program.
- This announcement follows OpenAI’s recent acquisition of OpenClaw, highlighting its focus on autonomous AI agents.
OpenAI has launched EVMbench to enhance the security of crypto tokens and smart contracts by evaluating AI capabilities. This initiative is part of a broader strategy that includes expanding Aardvark’s beta and acquiring OpenClaw.
Source (3.2)https://cryptobriefing.com/ai-security-benchmarking-system/?rand=59535
