Ethereum Foundation uses AI proxy red team to test ETH network and discover real vulnerabilities

Researchers from the Ethereum Foundation ProtocolSecurity team said in a blog post on Thursday that a series of AI agents have been deployed to test the software on which Ethereum relies, searching for vulnerabilities in encryption systems, protocol code and smart contracts. The vulnerabilities discovered by AI agents include a remotely triggered panic issue in the point-to-point layer libp2pgossipsub used by the Ethereum consensus client, which has been fixed and disclosed on Github as CVE-2026-34219. Researchers say AI agents are organized into specialized roles such as reconnaissance, search, fill in and verify that they are used to find potential attack paths, reproduce faults, and verify whether they are suitable for production code. The Ethereum Foundation said that instead of replacing security researchers, AI has changed the way it works, allowing teams to cover far more than manual review, but requires researchers to make more careful judgments about a large number of seemingly credible conclusions.