Tuesday, 9 December, 2025
Google Fortifies Chrome's AI Agent Security
Google unveiled a multi-layered security framework for Chrome's new Gemini-powered agentic features to prevent threats like prompt injection. The architecture includes a separate "User Alignment Critic" AI model to vet actions and "Agent Origin Sets" to restrict website access. Furthermore, users must give explicit consent for sensitive actions like financial transactions or accessing banking sites, prioritizing safety and control.
Read full story at TechCrunch