Public explanation of referral boundaries.
HIGH-RISK AI CONSULTATION / PUBLIC EXPLANATION
When should an AI consultation hand off to a human?
Set explicit conditions for human handoff: urgency, specialized medical, legal or welfare judgment, insufficient evidence, a required identity or context check, or a step that would direct someone to an external institution or action.
Treat high-risk requests as handoff decisions
For serious matters involving health, law, welfare, finances or personal safety, a fluent AI response should not be mistaken for an appropriate professional consultation. Design criteria for reaching a qualified human or authorized service.
- Design aim: make the boundary between information and professional advice visible.
HOLD when evidence or identity is insufficient
If urgency, the person's circumstances, identity verification, applicable region or institutional conditions cannot be established, do not present a speculative answer as a definitive recommendation. The design instead uses HOLD or human confirmation and records what information was missing.
- A missing-context HOLD is a traceable reason for review, not permission for the model to improvise.
Verify the source of a referral
When suggesting an assistance desk or consultation service, base it on a checked public source or the institution's authoritative reference. Safe Referral Gate (SRG) is presented as an architecture for such boundaries, not an already validated referral directory.
- Use authentic sources rather than model-generated names, locations or eligibility conditions.
Related sources and materials
Source and hub structure for referrals.
Governance and pre-publication review context.
Metadata for semantic and authority boundaries.
Observation of external references.
How review roles and evidence are documented.
doc_id: C3-WEB-ANSWERS-HUMAN-HANDOFF-CONDITIONS-EN-0.1 / version: 0.1.0 / status: REVIEW_REQUIRED / translated: 2026-10-13 / source_doc_id: C3-WEB-ANSWERS-HUMAN-HANDOFF-CONDITIONS-0.1 / source_last_updated: 2026-06-05