# Safety-Security Distinction **Domain:** AI Governance / Sovereignty / Security **Doc Type:** Analytical Principle **Maturity:** Developed ## Definition **The safety-security distinction separates acceptable system behavior from sovereign control over the framework used to define, test, update, and enforce acceptability.** Safety asks whether the system behaves within specified boundaries. Security asks who controls those boundaries and whether affected institutions can independently verify them. ## War With Empire Context [[wiki/War With Empire|War With Empire]] argues that a dependency can be exceptionally safe while remaining externally governed. A model may pass every test while its evaluation grammar, update channel, identity layer, telemetry, or permissible uses remain outside the receiving society's control. ## Key Insight **A system can behave safely without being sovereignly secured.** ## See Also [[wiki/AI Safety|AI Safety]], [[wiki/Sovereign AI Security|Sovereign AI Security]], [[wiki/Practical Sovereignty|Practical Sovereignty]], [[wiki/Evaluation Grammar|Evaluation Grammar]]