Security & governance

AI proposes, people approve, systems execute, and every action remains on record

These statements are architecture, not a slogan. Explore what Patty controls, how enforcement works, and why each outcome can be verified.

Foundation: the DARI protocol

An open protocol sits beneath every control domain. Authority begins with signed delegation, and each action ends with a verifiable receipt.

Explore DARI →
  1. Propose

    AI proposes the next action within the work context and policy boundary.

    Intent · target · expected impact
  2. Approve

    An authorized person reviews scope and risk, then signs the authority.

    Approver · scope · validity
  3. Execute

    The control kernel applies only approved authority through isolated tools and infrastructure.

    Policy decision · outcome
  4. Record

    The chain from proposal to outcome is sealed into a verifiable receipt.

    Provenance · signature · audit

Need to review the control architecture?

Contact us