GDI v3.0 — The Decision Architecture for Governed AI

GDI v3.0 open specification release. Introduces the Governed Decision Record (GDR), Confidence Threshold Model, Gate Taxonomy, and CI/CD Governance Integration pattern. Includes GDR JSON schema, gate classifier reference implementation (114 tests), IETF draft-farley-acta-signed-receipts conformance driver (5th conforming implementation), bilateral pattern documentation, and framework compatibility mappings for NIST AI RMF, ISO 42001, EU AI Act, ARAF v3.0, and AIGN OS. Apache 2.0.

Governed Decision Intelligence (GDI) v2.1

Interoperability release. GDI is now the 5th conforming implementation of
draft-farley-acta-signed-receipts (IETF), cross-verified against
TypeScript, Python, Rust, and Go implementations.

Added

• Conformance driver for ScopeBlind/agent-governance-testvectors --
  GDR embedded in signed receipt payload before execution:
  result_hash = sha256(JCS(GDR))
• Sample receipt: blocked decision with full GDR sealed before execution
• Bilateral pattern documentation for pre/post-execution receipt chaining
  in regulated deployments
• Fully populated GDR example, schema-validated against gdr.schema.json
• Insurance claims example README with four-gate walkthrough
• GDI v3 Core Specification

Conformance

All three checks pass: schema validation, Ed25519 signature verification
via @veritasacta/verify, and hash-chain integrity.

PR: ScopeBlind/agent-governance-testvectors#7
Discussion: microsoft/agent-governance-toolkit#276

Governed Decision Intelligence (GDI) v2.0

Core specification and reference implementation defining how AI deployment
decisions are structured, evaluated, and audited as governed artifacts.

This release establishes the decision layer as a formal system with explicit:

• Decision outcomes (go, conditional go, no-go, defer, escalate)
• Evidence requirements and completeness classification
• Ownership and approval structures with named accountability chain
• Audit and reconstruction paths via SHA-256 hashed GDR artifacts

GDI v2.0 extends prior work on Regulated Gate Decision Support (RGDS),
evolving decision gates into a complete decision governance system.

Includes

• GDI v2.0 Core Specification
• Governed Decision Record (GDR) JSON Schema
• Gate classifier reference implementation (114 tests, all passing)
• Framework compatibility documentation (NIST AI RMF, ISO 42001, EU AI Act,
  ARAF v3.0, AIGN OS)

Lineage

Regulated Gate Decision Support (RGDS) established decision gates as
structured governance artifacts for evaluating AI systems prior to deployment.
GDI v2.0 extends this foundation into a complete decision governance system,
defining how deployment decisions are structured, evaluated, and audited with
explicit evidence, ownership, and outcomes.