This is the policy layer that governs what AI assets are allowed to do — covering policy authoring, distribution across enforcement points, decisioning at central and local decision points, enforcement actions from explicit block to kill switch, and immutable audit of every decision. Policies are authored once centrally and enforced at every connected enforcement point. Policies include access control for humans and agents, kill switch, and others.
Policy authoring — low-code builder for common patterns, letting policy authors define conditions and actions without writing custom code.
Central distribution, local evaluation — policies authored once and pushed as bundles to every enforcement point, evaluated locally with no round trip; only conditions needing shared context reach back to the central decision point.
Enforcement actions — block or kill, scoped to a single session, device, agent, model, or everything at once.
Kill switch — applied as a policy targeting a specific configuration item, so it uses the same mechanism and audit path as every other policy rather than a special code path.
Threat response — a policy that watches for a specific threat type crossing a threshold you define, then automatically contains the affected agent without waiting for someone to notice and react.
Policy impact visibility — see which assets any given policy affects.
New MVP release of AI Control Tower's Policy Framework and kill switch.
Plugin Dependencies:
- September release of AI Control Tower for Enterprise (AI Native SKU) (sn_aict_foundation)