Transparency of Operation
AI systems must be capable of producing a complete, accurate account of their decision process on demand — not in summary, but in structural detail.
Human Authority Preservation
Systems must observe, structure, and expose — not decide, instruct, or override. Human authority is not delegated. It is architecturally preserved.
Externally Verifiable Limits
Behavioral constraints must be enforceable by an external authority — not assumed from design intent or operator self-report.
Auditability Without Intermediation
Governance authorities must be able to verify system behavior without routing that verification through the system's operator.
