RFC-034: Domains of Concern for Mediated Intelligence — 17. Security and Safety Considerations
AIGP Specification › RFC-034: Domains of Concern for Mediated Intelligence › 17. Security and Safety Considerations
← 16. Design Principles · Section index · 18. Summary →
17. Security and Safety Considerations
Incorrectly classifying a concern by industry rather than mediation constraint may produce unsafe or misleading evaluations.
Risks include:
- applying the wrong evidence requirements,
- assigning the wrong observer,
- ignoring uncertainty,
- treating unsupported model output as empirical,
- using inappropriate thresholds,
- missing required escalation,
- allowing autonomous action outside its envelope,
- generating false confidence in quality.
Systems implementing this RFC SHOULD prevent artifacts from being evaluated unless a Domain of Concern has been declared and the required evidence is available.