One exact action
Approval is bound to the requested action, target and parameters — not to a broad session.
Independent authorization and execution assurance
ExactPermit is developing an authorization, control and evidence layer for AI-agent actions, so an action proceeds only when it matches exactly what was approved.
Current status: validated local Safe Demo — not an unrestricted production release.
The control gap
AI agents can send messages, modify files, call APIs and operate business systems. ExactPermit places independent enforcement between an AI agent and real-world systems.
Approval is bound to the requested action, target and parameters — not to a broad session.
Missing or invalid human authority means no action for flows that require human approval.
Verification continues after execution to confirm that the result matched the permission granted.
Validated technical baseline
The figures below describe the controlled AI SHIELD V5.2 Safe Demo validated on 1 September 2026. They do not represent external certification or production approval.
Current stage
The next phase is productization: production hardening, selected integrations, user experience, limited pilot deployments and independent external security review.
ExactPermit
ExactPermit is preparing for structured conversations with strategic investors and potential design partners for one bounded, non-production pilot.
Business contact: maymon@exactpermit.com