External source
DIRECTArticle 14 · Human oversight
High-risk AI systems require effective human oversight appropriate to their risk, autonomy, and context of use.
- Status
- In force · consolidated 2026-07-27
- Source type
- Binding regulation
Methodology 1.0.0-rc.2 · Public calibration candidate · Not production active
The Alpha Standard · Crosswalks
Frameworks define what matters.
Alpha determines whether the enterprise can govern it.
The Alpha Standard maps material AI governance obligations, practices, and threats into a governed evidence and rating system. It makes performance comparable without claiming certification, compliance, or framework equivalence.
- Rated subject
- Enterprise governance system
- Analytical hierarchy
- 6 pillars · 24 subcategories · 48 requirements
- Rating family
- Alpha AI Governance Rating
The judgment layer
From public framework to comparable opinion.
A crosswalk is useful only when it preserves applicability, evidence, and judgment. Select a source to inspect the path Alpha exposes.
Alpha requirement
PRIMARY HOMEIntervention, override & shutdown
Material AI and agents must have tested, authorized intervention paths that preserve containment, attribution, and safe restart.
- Pillar
- Monitoring & Improvement
- Cadence
- Quarterly and event-driven
Evidence and test
OPERATING EVIDENCETrace an intervention exercise
Inspect authority, shutdown, credential revocation, downstream containment, state integrity, and authorized restart.
- Public visibility
- Low
- Score anchor
- 0 Absent → 4 Assured
Alpha judgment
HUMAN APPROVEDRating symbol, outlook, and rationale
The calculated result may be constrained by a critical-pillar ceiling. Missing applicable evidence produces NR, never an inferred pass.
- Output
- AAA-D or NR
- History
- Immutable observation
Crosswalk mappings indicate relevance and possible applicability. They do not constitute legal advice, certification, safe harbor, compliance, or framework equivalence. Source records reviewed 2026-08-04.
What Alpha assesses
One governance question. Six scored pillars.
Each pillar contains four equal-weight subcategories and eight requirements. A strong average cannot erase a severe failure in a critical pillar.
PillarBoard questionWeight
01Leadership & Accountability
Who is responsible for AI, and can the board hold them accountable?
20%02Safety, Security & Resilience
Can AI operate safely and withstand attack, failure, and disruption?
22%03Data, Privacy & Transparency
Is data protected, and are AI uses, decisions, and claims clear?
15%04People & Rights
Are people treated fairly, protected from harm, and able to challenge decisions?
14%05Compliance & Third Parties
Are legal duties, vendors, models, and external dependencies governed?
15%06Monitoring & Improvement
Can the enterprise detect problems, intervene, correct them, and learn?
14%Framework library
Mapped by obligation, practice, and threat.
The library is organized by source family and status. Every public record carries a source, review date, and applicability boundary.
EU
EU AI Act
Role- and risk-based legal obligations mapped to entity-specific applicability.
In force · consolidated 2026-07-27US
NIST AI RMF
Govern, Map, Measure, and Manage outcomes traced to Alpha requirements and tests.
AI RMF 1.0 under revisionISO
ISO/IEC 42001
AI management-system clauses mapped to evidence of authority and operating effectiveness.
Published · 2023OECD
OECD AI Principles
Trustworthy-AI principles connected to accountability, rights, transparency, and resilience.
Updated · 2024ATLAS
MITRE ATLAS
AI adversary tactics, techniques, and mitigations linked to controls and test evidence.
MaintainedOWASP
LLM and agentic security guidance
Application and agentic weaknesses mapped to safety, ecosystem, and intervention controls.
Maintained · 2026 materials reviewedSOC
SOC 2
Assurance evidence may support Alpha requirements but does not establish AI governance alone.
Supporting evidence27001
ISO/IEC 27001
Information-security evidence is evaluated within scope, freshness, and relevance limits.
Published · 2022; amended 2024Source library last reviewed .
One rating, two evidence bases
Same standard. Different evidence perimeter.
Public and verified observations use the same analytical structure but remain separate. Private evidence cannot enter, alter, or be inferred from a public rating.
Public Information
Attributable public evidence
Filings, policies, regulator records, incidents, litigation, standards activity, and attributable reporting available by the cutoff date.
- Normal maximum
- A
Verified Assessment
Evidence verified by Alpha
Board records, inventories, tests, change logs, vendor registers, incidents, and corrective actions within a signed assessment perimeter.
- Available scale
- AAA-D
Boundary of opinion
What the Alpha Standard does not replace.
An Alpha Governance Rating is an independent governance opinion. It is not a legal conclusion, product certificate, audit opinion, or substitute for accountable decision-making.
- 01Legal advice or an entity-specific compliance determination
- 02EU conformity assessment or notified-body certification
- 03ISO certification, SOC 2, or a financial-statement audit
- 04Product safety testing, red teaming, or penetration testing
- 05Regulator, court, auditor, insurer, management, or board judgment
Alpha Ratings
Examine the judgment system.
Review the methodology, evidence boundaries, and framework mappings with Alpha.