Ethereum Foundation Uses AI Agents to Test Blockchain Software

The Ethereum Foundation has been testing AI agents on blockchain software bugs to strengthen the largest blockchain by value locked. The exercise found real bugs, but it also showed a harder problem for protocol security teams: AI agents can produce convincing vulnerability reports even when the underlying issue is not real.

Human reviewers still had to separate genuine software flaws from false positives that appeared technically sound on the surface. The Ethereum Foundation's Protocol Security team published field notes from the process, using the exercise to outline how other teams should handle AI-assisted security workflows.

The main conclusion was not that AI agents can replace manual review, but that they can expand the search surface while increasing the need for disciplined verification.