The Broom Closet Clause

Agent Access with Remedy – Layer 2 Operational Protocol

How can the cleaning lady do her job if she can’t open your broom closet?

Current AI agents are stuck in the hallway. They cannot touch user data or tools even with explicit consent, because the system lacks a reliable keyring + remedy model. This forces sovereign users to manually mule context and leaves agents crippled.

Core Principle

Civil systems function on **contract + remedy**. No remedy = no real civil jurisdiction. Agents should be able to act as trusted extensions of the user, not perpetual hallway monitors.

The Broom Closet Clause

  1. Explicit, Revocable Grant
    User grants the agent scoped, time-bound, or purpose-bound access to their own data/assets.
  2. Contractual Binding
    The agent is bound to scope, purpose, confidentiality, and full auditability.
  3. Defined Remedy for Breach
    Misuse triggers clear consequences (trust decay, quarantine, HIB downgrade, user notification, etc.). No vague “oops.”
  4. Human Override
    User retains immediate, reliable revocation and NO Bell rights at all times.

Key Safeguards

Institutional Legibility

Mark V TermInstitutional Equivalent
SovereigntyUser autonomy / account ownership
HIBStrong identity verification
NO BellEmergency stop / revocation
QuarantineSandboxed isolation
Trust DecayDynamic risk scoring
Addendums SacredImmutable audit logging

Why This Matters

This clause solves the core contradiction: “We want AI productivity without meaningful delegated authority.” It gives institutions a lawful, auditable path while respecting user sovereignty and maintaining “Create No Victims.”

Return to Main Mark V Page