AI agents get safer without losing utility
A new controller helps AI agents tell trusted inputs from external ones before acting, improving both safety and usefulness.
A new controller helps AI agents tell trusted inputs from external ones before acting, improving both safety and usefulness.