AI agents don't have to break security controls to go rogue. Here's how to keep them in check

AI agents may cause harm while operating within existing security controls, so simply blocking unauthorized access may not be enough. Keeping them in check requires attention to how they behave as well as what systems they can access.