AlphaWhat we measure

The public standard for AI governance quality.

The Alpha Governance Standard evaluates the enterprise environment around AI, not individual models in isolation. The Alpha 12 defines what Alpha measures, AGR expresses governance quality, and GMI expresses how material AI governance is to the company.

Three related outputs, one standard.

Alpha separates the standard, the rating opinion, and the materiality view so decision-makers can distinguish governance quality from governance exposure.

AGS

Comprises the Alpha 12 and defines what Alpha evaluates across governance structure, controls, disclosure, resilience, stakeholder impact, and corrigibility.

AGR

Expresses governance quality as an independent rating opinion on the Alpha scale.

GMI

Expresses governance materiality by assessing operational dependency, strategic centrality, risk exposure, and stakeholder impact.

Governance quality is assessed across the Alpha 12.

The Alpha 12 defines the twelve dimensions Alpha evaluates. The rating file applies profile-specific analysis for developers, deployers, and hybrid companies without exposing exact scoring mechanics.

D1

Leadership and Accountability

Whether AI governance has named board and executive ownership, clear mandates, reporting cadence, and consequences for failure.

Evidence considered

  • -Board charter
  • -Executive owner
  • -Committee reporting

D2

AI Safety and Robustness

How the organization tests, constrains, monitors, and responds to material AI system failures and unsafe behavior.

Evidence considered

  • -Testing records
  • -Incident playbooks
  • -Deployment safeguards

D3

Transparency and Disclosure

The quality and consistency of AI governance disclosures to boards, investors, regulators, customers, and the public.

Evidence considered

  • -Annual reports
  • -Proxy disclosures
  • -Risk disclosures

D4

Data Ethics and Privacy

Governance of the data used by AI systems, including consent, provenance, minimization, privacy, and data supply-chain controls.

Evidence considered

  • -Data policies
  • -Privacy reviews
  • -Data provenance controls

D5

Governance Infrastructure

The operating system of AI governance: policies, committees, escalation paths, audit mechanisms, and enterprise-risk integration.

Evidence considered

  • -Policy framework
  • -Escalation paths
  • -Audit coverage

D6

Fairness and Bias Mitigation

Whether bias and adverse impacts are identified, measured, remediated, and monitored after deployment.

Evidence considered

  • -Bias testing
  • -Remediation records
  • -Monitoring reports

D7

Societal, Workforce, and Environmental Impact

How the organization governs AI impacts on workers, customers, communities, and environmental footprint.

Evidence considered

  • -Workforce plans
  • -Impact assessments
  • -Stakeholder reporting

D8

Regulatory and Compliance Posture

Readiness for current and emerging AI governance requirements without reducing the rating to a compliance checklist.

Evidence considered

  • -Regulatory mapping
  • -Control attestations
  • -Supervisory engagement

D9

Cybersecurity and AI Resilience

Controls for AI-specific security, adversarial misuse, model or agent compromise, and operational resilience.

Evidence considered

  • -Security testing
  • -Access controls
  • -Resilience exercises

D10

Third-Party and Supply Chain Governance

Governance of AI vendors, model providers, data suppliers, integration partners, and downstream accountability.

Evidence considered

  • -Vendor reviews
  • -Contract controls
  • -Third-party monitoring

D11

Stakeholder Impact Layer

How the organization identifies, measures, discloses, and remediates AI impacts on affected stakeholders, including customers, workers, communities, and other external parties.

Evidence considered

  • -Affected-party mapping
  • -Impact assessments
  • -Remediation records

D12

Corrigibility Assessment

Whether AI systems can be corrected, overridden, constrained, or shut down when governance requires it, with documented human control and board-visible reporting.

Evidence considered

  • -Override mechanisms
  • -Shutdown protocols
  • -Corrigibility event records

What Alpha does - and does not - measure.

We do measure

  • - The governance environment around AI deployment
  • - Evidence of oversight, controls, and disclosure
  • - How governance performs under pressure, not just policy design
  • - AI governance materiality through GMI exposure factors

We do not measure

  • - Individual model certification or technical safety in isolation
  • - Brand or marketing claims about AI use
  • - Forward-looking financial performance
  • - Exact scoring recipes that would allow ratings to be gamed

Request the published AGR methodology paper, v2.4.