P1
Human dignity first
AI systems must not degrade the rights, safety, or standing of the people they affect.
Trust is impossible when a system's design assumes that individuals must bear the cost of institutional error.
Pillar 1 · The Standard
The Ethos Trust Standard
A versioned, citable public standard for demonstrating — not declaring — trust in AI systems. Six principles, eight controls, three conformance levels. Mapped to the NIST AI Risk Management Framework and ISO/IEC 42001, and licensed for public use.
- Version
- v1.0.0
- Status
- Public Review
- Effective
- 2026-01-01
- Last updated
- 2026-07-21
Foundational Standard
Why Standards Matter
Across cybersecurity, quality management, accounting, engineering, and safety, organizations rely on recognized standards to create consistency, accountability, and public confidence.
Artificial intelligence should be no different.
The Ethos Trust Standard™ provides a transparent framework for demonstrating trustworthy AI through documented governance, measurable controls, and verifiable evidence — not marketing claims or self-declarations.
Trust is not a feature. It is the result of transparent governance, measurable evidence, and continuous accountability.
As artificial intelligence becomes embedded in healthcare, education, government, finance, critical infrastructure, defense, and everyday life, trust can no longer depend on reputation alone.
Organizations must be able to demonstrate:
- what was built
- how it was governed
- what evidence supports its claims
- how accountability is maintained over time
The Ethos Trust Standard™ creates a common language for demonstrating those commitments in a way that can be understood by executives, implemented by technical teams, reviewed by auditors, and trusted by the public.
Evidence-Based
Every claim should be supported by verifiable documentation.
Transparent
Governance should be understandable and inspectable.
Accountable
Organizations remain responsible for AI outcomes.
Versioned
Standards evolve through documented revisions.
Public Trust
Confidence is earned through openness and evidence.
Independent Review
Designed to support objective assessment.
The Ethos Trust Standard™ is maintained as a living public standard. New versions are developed through research, expert review, public feedback, and governance oversight to ensure the framework evolves alongside advances in artificial intelligence while preserving transparency, stability, and public trust.
§1 · Principles
Six principles
The non-negotiable commitments any conforming system must uphold.
P2
Evidence over assertion
Claims about fairness, accuracy, or safety must be backed by verifiable, reproducible evidence.
Public trust is not established by policy documents; it is established by inspectable proof.
P3
Meaningful transparency
Disclosures must be legible to affected people, not only to developers or regulators.
Transparency that requires expertise to interpret is not transparency; it is compliance theater.
P4
Contestability
People subject to automated decisions must have a real path to challenge, appeal, and correct them.
A system that cannot be contested cannot be corrected, and cannot earn trust over time.
P5
Proportional oversight
Human oversight, review cadence, and controls must scale with the stakes of the decision.
Low-risk assistants and life-altering decision systems cannot be governed by the same posture.
P6
Continuous accountability
Trust is a public, ongoing commitment — not a one-time certification.
Systems change. Populations change. Accountability must be re-attested, not archived.
§2 · Controls
Eight controls across five domains
Each control specifies a public requirement and the evidence that satisfies it. External framework mappings are provided where they exist.
| ID | Domain | Requirement | Evidence | Maps to |
|---|---|---|---|---|
| C-GOV-01 | Governance | Publish a named accountable owner and escalation contact for the system. | Public registry entry with owner, role, and contact channel. | GOVERN 2.1 · ISO/IEC 42001 §5.3 |
| C-GOV-02 | Governance | Maintain a versioned model card and impact assessment reviewed at least annually. | Versioned artifact linked from the public registry entry. | GOVERN 1.4 · ISO/IEC 42001 §6.1 |
| C-MAP-01 | Map | Document intended use, prohibited uses, and populations reasonably foreseen to be affected. | Public use-scope statement in the registry entry. | MAP 1.1 |
| C-MEAS-01 | Measure | Report performance metrics disaggregated by protected and affected subgroups. | Subgroup performance table in the trust report. | MEASURE 2.11 |
| C-MEAS-02 | Measure | Publish incident, error, and override rates on at least a quarterly cadence. | Time-series data in the public trust dashboard. | MEASURE 3.1 |
| C-MAN-01 | Manage | Provide a contestability channel with a documented response SLA. | Public appeals process and quarterly appeals outcomes. | MANAGE 2.3 |
| C-MAN-02 | Manage | Support independent review by an accredited third party at least every 24 months. | Independent review report published in the registry. | ISO/IEC 42001 §9.2 |
| C-DISC-01 | Disclosure | Notify affected people, in plain language, when an automated decision materially affects them. | Sample notice archived with the registry entry. | — |
§3 · Conformance
Three levels of demonstrated trust
Operators may declare, verify, or independently review conformance. Each level adds public scrutiny — not additional paperwork.
Level 1
Declared
The operator publicly attests to conformance and publishes the required evidence.
- Signed public attestation by an accountable owner.
- All governance disclosures published in the registry.
- Annual re-attestation.
Level 2
Verified
Ethos staff reviewers verify the published evidence against the standard.
- All Declared requirements.
- Staff review of measurement, disclosure, and contestability controls.
- Public reviewer note attached to the registry entry.
Level 3
Independently Reviewed
An accredited third-party reviewer certifies conformance and publishes findings.
- All Verified requirements.
- Independent reviewer engagement within the last 24 months.
- Public review report and reviewer accreditation ID.
§4 · Versioning
Changelog
The Standard is versioned semantically. Every change is dated, described, and preserved.
v1.0.02026-07-21Public Review
Initial public review release. Six principles, eight controls across Governance / Map / Measure / Manage / Disclosure, three conformance levels.
Cite this standard
Ethos AI Registry. The Ethos Trust Standard, v1.0.0. 2026-07-21. https://ethosairegistry.org/standard
License & use
Released under CC BY 4.0. Governments, researchers, and operators may adopt, adapt, and reference the Standard with attribution. The JSON schema is a stable, machine-readable artifact.