Google AI Finds 13-Year Chrome Sandbox Bypass Vulnerability

Google's AI vulnerability agent, powered by its Gemini security LLM model, has found a significant security flaw in Chrome. This AI tool uncovered a sandbox escape bug that had gone undetected for over 13 years. What's interesting is this bug could have allowed Chrome to access local files. Google developed this AI agent through collaborations, including with Project Zero and DeepMind. The company is now adding restrictions to the AI's detection capabilities and plans to use AI more for validating, triaging, and fixing bugs. The good news is the bug was fixed before it could compromise Chrome's security. This matters because it shows how AI is being used to find and fix long-standing vulnerabilities.