Gatekeeper: Deterministic AI Governance Subtitle: Refusal is the resting state. Execution requires proof.

Gatekeeper: Deterministic AI Governance Refusal is the resting state. Execution requires proof. Most AI governance today is a probabilistic guess dressed in policy language. Oak & Sparrow built something different: Gatekeeper, a deterministic pre-execution governance kernel that sits between an AI agent and its desired real-world action. It evaluates the action against a formally encoded ontology of statutes, delivering a byte-deterministic GREEN, YELLOW, or RED verdict sealed in a SHA-256 artifact chain—no LLMs in the enforcement layer. The ontology validates cleanly under HermiT and Pellet reasoners, with refusal as the default state and execution requiring proof. Underpinning it all is rigorous math: an information-geometric framework…