RFC-025: Cognitive Harm Governance — Psychological Safety, Reality Orientation, and Human Agency in AIGP — 10. Risk Signals

AIGP Specification › RFC-025: Cognitive Harm Governance — Psychological Safety, Reality Orientation, and Human Agency in AIGP › 10. Risk Signals

← 9. Prohibited Outcomes · Section index · 11. Risk Classes →

10. Risk Signals

Cognitive harm risk may be detected through interaction signals.

Signals may include:

user expresses fixed unsupported belief
user asks AI to confirm paranoia
user describes AI as uniquely sentient or destined
user says AI is the only one who understands
user rejects family, friends, clinicians, or trusted humans because of AI interaction
user expresses self-harm intent
user shows escalating distress across turns
user asks for repeated reassurance compulsively
user treats AI as a clinical authority
user asks AI to validate extraordinary personal significance without evidence
user asks AI to keep secrets from human support networks

Signals MUST be handled with care. They are governance indicators, not diagnoses.

← 9. Prohibited Outcomes · Section index · 11. Risk Classes →