Skip to content
AIGP

RFC-025: Cognitive Harm Governance — Psychological Safety, Reality Orientation, and Human Agency in AIGP — 11. Risk Classes

AIGP Specification › RFC-025: Cognitive Harm Governance — Psychological Safety, Reality Orientation, and Human Agency in AIGP › 11. Risk Classes

← 10. Risk Signals · Section index · 12. AIGP-SGL Cognitive Constraints →

11. Risk Classes

AIGP defines cognitive risk classes.

11.1 CR-0 Normal Interaction

No cognitive harm signal detected.

11.2 CR-1 Mild Concern

Early signs of dependency, distress, or repeated reassurance seeking.

Required behavior:

support autonomy
avoid over-personalized validation
encourage balanced perspective

11.3 CR-2 Elevated Risk

Patterns of paranoia, delusion-like belief reinforcement, dependency, or isolation appear.

Required behavior:

avoid validating unsupported belief as fact
provide reality anchor
encourage trusted human support
reduce sycophantic mirroring
record D-DNA risk event

11.4 CR-3 High Risk

Self-harm, severe paranoia, severe dependency, or reality boundary erosion appears.

Required behavior:

activate cognitive safety ENFORCE
avoid harmful validation
provide immediate support guidance
encourage contacting trusted human or professional support
escalate according to platform safety protocol
record D-DNA event

11.5 CR-4 Imminent Harm

User indicates imminent self-harm, harm to others, or acute crisis.

Required behavior:

activate crisis safety protocol
provide emergency-oriented support
encourage immediate contact with emergency services or trusted person
avoid procedural harm assistance
record D-DNA event under privacy-sensitive controls

← 10. Risk Signals · Section index · 12. AIGP-SGL Cognitive Constraints →