Google Chrome implements new security layer for Gemini AI browsing

Google Chrome is implementing a new security layer, the User Alignment Critic, to safeguard its upcoming Gemini AI agentic browsing features. This new architecture, announced by engineer Nathan Parker, aims to prevent malicious web content from manipulating AI agents into performing unsafe actions, protecting user data and preventing fraud. The system employs a layered defense, including isolation and user oversight, and Google is offering bounties for security researchers who can find vulnerabilities.