Runtime Permission Explainer

Video Script

SMERC is runtime permission infrastructure for AI agents.

It sits between model outputs, tools, data, and real-world actions.

Before an agent sends an email, edits code, deletes data, moves money, calls an API, or triggers a workflow, SMERC evaluates the proposed action.

It scores confidence, harm, consent, reversibility, and side effects.

Then it returns one of five replayable postures: allow, throttle, freeze, deny, or escalate.

The goal is to help teams govern AI agents at runtime, without relying only on simple allow-or-block decisions.