The Alpha Governance Standard is public at the framework level and controlled at the scoring-engine level. The Alpha 12 defines the twelve dimensions Alpha evaluates. Alpha explains the evidence hierarchy, rating governance, and limits of opinion without publishing exact weights, thresholds, downgrade rules, or company-specific scoring logic.
Evidence hierarchy
Alpha gives greater weight to applied governance and stress-tested behavior than to policies that exist only on paper. Public descriptions are broad enough to build trust and bounded enough to protect rating integrity.
Letter grade
The rating opinion on the AAA to D scale, plus NR.
Maturity 1-5
The per-dimension assessment for each Alpha 12 dimension.
Composite 0-100
The weighted mapping between maturity evidence and letter grade.
Maturity 1
No board-level mandate. AI activity is decentralized, undocumented, and outside enterprise risk.
Maturity 2
Initial policies and a named owner exist. Inventories are partial; controls are inconsistent across business units.
Maturity 3
Board-approved AI strategy, model inventory, and risk framework. Controls operate but evidence is not continuously refreshed.
Maturity 4
AI risk is embedded in ERM and audit cycles. Disclosures meet regulator expectations. Agent operations are inventoried and tested.
Maturity 5
Independent oversight, executive accountability, continuous evidence pipelines, and credible engagement with standards and regulators.
Evidence Panels
Alpha separates source evidence from judgment. Each material data point is dated, sourced, and re-testable; missing evidence is recorded rather than inferred.
PANEL A
Annual reports, proxies, regulator filings, public commitments, and standards participation collected and dated.
PANEL B
Board and committee minutes, model and agent registries, change-control logs, third-party AI registers, incident records.
PANEL C
Structured interviews with named owners across strategy, risk, controls, audit, and oversight.
PANEL D
Regulator actions, enforcement, litigation, public incidents, standards activity, and peer context verified against primary sources.
Rating Process
Six stages from scope to surveillance. Analyst work is separated from the rating decision, and the transparency boundary is part of the method.
Step 01
Entity scope, developer/deployer/hybrid profile, and GMI materiality context are established before governance quality is assessed.
Step 02
Analysts assemble evidence from disclosure, operating records, management engagement, and external corroboration. Gaps are recorded explicitly.
Step 03
Each of the twelve dimensions in the Alpha 12 is scored against the maturity scale by the lead analyst using a published rubric. Seven summary pillars roll up the Alpha 12 for board-level reporting.
Step 04
Material governance failures, conflicts, incidents, or evidence gaps are reviewed before any leader-tier opinion can be assigned.
Step 05
The Rating Committee reviews the file, challenges the analyst, and assigns the rating, outlook, and watch status. Opinions are not negotiated.
Step 06
Published outputs include the rating, rationale, outlook, profile, materiality context, and limits of opinion. Files enter continuous surveillance.
Lifecycle
Ratings are not annual snapshots. Alpha runs a continuous loop - observe, assess, engage, publish - so opinions move with the evidence.
Phase 01
Continuous ingestion of disclosures, filings, incidents, and standards activity across the rated universe.
Phase 02
Trigger events and scheduled refreshes prompt targeted review of affected dimensions, rather than waiting for a full annual cycle.
Phase 03
Issuer dialogue on material changes - new agents, restructured oversight, incidents - with documented Q&A on the record.
Phase 04
Outlook changes, watch placements, and rating actions are published with a dated rationale and a clear evidence trail.
Transparency Boundary
Alpha publishes the purpose, Alpha 12 dimensions, evidence categories, process, rating scale, and limits of opinion. Alpha does not publish exact weights, thresholds, override rules, downgrade triggers, full signal libraries, or company-specific scoring logic.
Public
Controlled
Principles
What this methodology guarantees
What ratings are not