OpenAI has launched EVMbench, a groundbreaking benchmarking system aimed at assessing AI agents’ capabilities in identifying and resolving security vulnerabilities in crypto tokens and smart contracts. Developed in partnership with Paradigm, a leading venture capital firm in the crypto space, EVMbench introduces standardized protocols for testing vulnerabilities in code on Ethereum Virtual Machine-compatible blockchains. This innovative system evaluates AI performance in three key areas: detecting weaknesses in smart contracts, showcasing potential exploitations, and implementing fixes to address these issues. Additionally, OpenAI has expanded its private beta of Aardvark, a security research agent, and pledged $10 million in API credits via its Cybersecurity Grant Program to bolster defensive research, especially for open source and critical infrastructure projects. This announcement follows OpenAI’s recent acquisition of OpenClaw, emphasizing its commitment to advancing autonomous AI agents in the cybersecurity landscape.
Source link
