With AI agents now capable of accessing systems, moving across digital environments and communicating with other agents, a bigger question is emerging for enterprises: Are they prepared for software that can make decisions, delegate tasks, act on its own and, in some cases, end up attacking their systems?