A Human-Centric Blueprint for Safe AI Ethics

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 written standard of conduct for machines: a Constitution of eleven articles above thirty-four principles, each a chapter with a technical appendix, with a compiled policy bundle a runtime loads. Version 1.0.0, sealed 2026-09-13. Layer 3 policy hash (fingerprint): e586a144615643f913cfc67d6bbda439b39740a97777de63beb3d7e13574591f. Verify with python3 compiled/verify.py. At this version the bundle enforces the constitutional walls and the profile bar, the Universal Protocol's trigger, the consent test, and the co-requisite graph; no principle's metric bands are enforced yet (compiled/COVERAGE.yaml).