Ethereum Foundation AI Finds Bugs, Human Review Key

The Ethereum Foundation is using artificial intelligence to find security vulnerabilities in its client software. An AI agent successfully identified a flaw in the Gossipsub message propagation protocol. This vulnerability could cause validator nodes to go offline. It has since been patched and assigned CVE-2026-34219. Here's the thing: while AI can find issues, it struggles to tell real vulnerabilities from false alarms. The AI might generate plausible but non-existent problems, needing human researchers to validate them. The Foundation notes that AI is better at single-code issues and less effective at complex attack chains. Looking ahead, the Ethereum Foundation plans to use AI to help generate potential attack paths. This will be combined with manual and automated testing to improve how they find and confirm vulnerabilities. This work aims to make the Ethereum network more secure.