Terminal agents are safer and more controllable than device agents due to the adoption of the principle of least privilege, making them more suitable for enterprise scenarios.