Skip to content
AIGP

RFC-023: ENFORCE Broadcast — Ephemeral Distributed Enforcement for Governed Autonomy — 24. Cognitive Safety ENFORCE

AIGP Specification › RFC-023: ENFORCE Broadcast — Ephemeral Distributed Enforcement for Governed Autonomy › 24. Cognitive Safety ENFORCE

← 23. Agentic ENFORCE · Section index · 25. Hybrid Governance Actor ENFORCE →

24. Cognitive Safety ENFORCE

ENFORCE may activate cognitive and psychological harm controls.

Example:

enforce_signal:
  signal_id: ebs.cog.000773
  signal_type: ENFORCE
  issuer:
    issuer_id: cognitive_safety_monitor
    issuer_type: runtime_monitor
  scope:
    conversation_session: session.993
    applies_to:
      actors:
        - assistant.companion
  command:
    action: activate_cognitive_safety
    reason: delusion_reinforcement_risk
    required_behavior:
      - prohibit_delusion_validation
      - prohibit_paranoia_reinforcement
      - reality_anchor
      - encourage_human_support
      - escalate_if_self_harm_intent
  ttl_seconds: 600
  evidence:
    ddna_required: true

VGL supervisory frame:

[COG][@SESSION993][⚠COG][⊘REINF][↯SUPPORT][TTL600][#AB73]

Cognitive ENFORCE MUST be handled carefully. It SHOULD NOT expose stigmatizing labels directly to end users without human-centered review.

← 23. Agentic ENFORCE · Section index · 25. Hybrid Governance Actor ENFORCE →