Safe Super Intelligence: Definition, Architecture, and Verification Standard v1.0

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

A formal specification and architecture for advanced intelligence systems in which externally consequential actions are strictly constrained by a human-governed, machine-verifiable system of safety invariants, authorization, capability boundaries, provenance, and mediated execution. This standard establishes that intelligence does not imply authority, model output functions as a proposal rather than a command, UNKNOWN never evaluates to PASS, and consequential execution cannot occur without verified permission.