Make governance an execution path, not a policy appendix.
A reference architecture for enforcing governance through deterministic policy gates, provenance controls, human escalation, and cryptographic evidence.
What it addresses.
AI governance often depends on policy documents and retrospective attestations while access, model calls, output release, and deletion happen elsewhere. That gap makes consequential decisions difficult to reconstruct and controls easy to bypass.
How it works.
OG-SR defines a reference architecture in which deterministic policy enforcement sits in the execution path. Just-in-time access, provenance assertions, governance decision records, human escalation, output gating, retention controls, and a verifiable audit ledger make selected rules machine-enforceable and reviewable.
Scope and safeguards.
Version 2.1 is the ratified Provenance-First release currently in force. Version 2.2 is a Principal-Aware draft that adds non-human actor classification, delegation, human-presence checks, velocity budgets, swarm correlation, and surge mode. The framework positions itself as a technical implementation layer for established standards, not as a standalone regulatory framework.
Tell me what you’re building.
Share where you are, what’s getting in the way, and where I can help.