Permissions, network boundaries, tool allowlists, rate limits, and approval gates reduce what a compromised or mistaken model can do. Controls should be enforced by the surrounding system rather than by prompt instructions alone.
Capability control limits which actions, information, or resources an AI system can access.
Permissions, network boundaries, tool allowlists, rate limits, and approval gates reduce what a compromised or mistaken model can do. Controls should be enforced by the surrounding system rather than by prompt instructions alone.