Principle 1
Human dignity first
AI systems must protect the rights, agency, and safety of the people they affect — especially those with least power in the transaction.
Core Pillar
Ethics & Public Trust
Compliance proves an organization followed a rule. Trust proves the public can verify it. The Ethics & Public Trust pillar of the Ethos AI Registry gives organizations the principles, tooling, and public infrastructure to be trusted — not merely compliant — with the AI systems they deploy.
- Public by default
- Evidence is published, not filed away.
- NIST-aligned
- Mapped to NIST AI RMF and ISO/IEC 42001.
- Contestable
- Every consequential decision has an appeal path.
- Continuous
- Re-audited and re-reported, not one-and-done.
Ethical AI Principles
Six principles the rest of the pillar is built on.
These principles are not aspirations. Each one is translated into a concrete requirement in the Registry, the audit methodology, or the public trust indicators.
Principle 2
Evidence over assertion
Claims of safety, fairness, or accuracy must be backed by publishable evidence — not marketing language or private assurances.
Principle 3
Meaningful transparency
Disclosure exists to be understood. Model cards, data lineage, and known failure modes are written for the public, not just auditors.
Principle 4
Contestability
Anyone materially affected by an AI decision must have a clear, timely path to challenge it and reach a human.
Principle 5
Proportional oversight
The higher the stakes — health, liberty, benefits, employment — the stricter the review, monitoring, and shutdown authority.
Principle 6
Continuous accountability
Governance is not a launch checklist. Systems are re-audited, incidents are logged, and public commitments are tracked over time.
Capabilities
The infrastructure of trustworthy AI.
Twelve capabilities that together let an organization earn — and defend — public trust in its AI systems.
Transparency Center
What it is. A single public destination where an organization publishes its AI systems, audit scores, model cards, and known limitations.
How it works. Powered by the Ethos Registry schema so disclosures are structured, comparable, and citable by regulators and press.
- Public system profiles with NIST-aligned audit scores
- Model cards and data-provenance summaries
- Change log of material updates and re-audits
Public Trust Dashboard
What it is. A living dashboard that turns audits, incidents, and community reports into a public trust signal for each deployed system.
How it works. Aggregates transparency score, incident volume, response time, and independent review status into indicators anyone can read.
- Per-system trust indicators
- Sector-level trust benchmarks
- Downloadable trust reports for boards and procurement
Human Oversight
What it is. Documented human-in-the-loop and human-on-the-loop controls for consequential AI decisions.
How it works. Registry submissions must declare who can override, pause, or shut down the system, and how contested decisions escalate.
- Named accountable owner per system
- Override, pause, and rollback procedures
- Escalation and appeal pathways for affected people
Responsible Innovation
What it is. A pre-deployment discipline that pairs product ambition with harm anticipation, red-teaming, and impact assessment.
How it works. Teams complete an Ethos pre-deployment questionnaire covering intended use, foreseeable misuse, affected populations, and rollback plan.
- Pre-deployment impact assessments
- Red-team findings and mitigations
- Go / no-go decision records
Community Reporting
What it is. A public channel for the people affected by an AI system to report harm, bias, or malfunction with evidence.
How it works. Structured community reports (category, description, evidence link) are triaged, published, and linked to the responsible system.
- Verified community reports on system profiles
- Bias-category taxonomy for trend analysis
- Response and resolution status tracking
Governance Board Toolkit
What it is. A working kit for boards, AI governance committees, and Chief AI Officers to run defensible oversight.
How it works. Board charters, meeting agendas, risk registers, and KPI templates aligned to NIST AI RMF and ISO/IEC 42001.
- AI governance committee charter template
- Quarterly board review agenda
- AI risk register and KPI pack
Public Commitments
What it is. Signed, dated statements an organization makes about how it will build, deploy, and be held accountable for AI.
How it works. Commitments are published on the organization's Transparency Center and tracked against evidence in the Registry.
- Standard commitment framework
- Public commitment page per organization
- Annual progress attestation
Independent Reviews
What it is. External, non-conflicted reviews of high-stakes AI systems by qualified auditors and civil-society experts.
How it works. Reviewers apply the Ethos audit methodology and publish findings, limitations, and recommended remediations openly.
- Independent review reports
- Reviewer qualification standard
- Remediation tracking to closure
Transparency Reports
What it is. Periodic public reports summarizing AI use, incidents, community reports, and governance actions.
How it works. Standardized annual and quarterly formats so reports are comparable across organizations and years.
- Annual AI transparency report template
- Quarterly incident and response summaries
- Year-over-year trust indicator movement
Trust Indicators
What it is. A concise, public-facing set of signals — like nutrition labels — that let anyone quickly assess a system's trustworthiness.
How it works. Indicators include transparency score, human oversight status, independent review status, incident response time, and appeal availability.
- Embeddable trust badge for organizations
- Machine-readable trust signals feed
- Public documentation of each indicator
Responsible Procurement
What it is. Procurement language and evaluation criteria that let public agencies and enterprises buy AI responsibly.
How it works. Model RFP clauses, vendor questionnaires, and scoring rubrics mapped to Ethos trust indicators and NIST AI RMF.
- Model AI procurement clauses
- Vendor AI risk questionnaire
- Bid-evaluation scoring rubric
Ethical AI Principles
What it is. The six-principle foundation every capability above is built on — a public standard organizations can adopt.
How it works. Principles are versioned, licensed for reuse, and translated into concrete registry, audit, and reporting requirements.
- Adoptable principles statement
- Mapping to NIST AI RMF and ISO/IEC 42001
- Public adopter list
Trust vs. compliance
Compliance is the floor. Trust is what the public actually rewards.
Regulation tells an organization what it must do. Trust is what happens when the same organization can be independently verified doing it — repeatedly, in public, over time. These capabilities are the shortest path from one to the other.
| Compliance posture | Trust posture (the Ethos standard) |
|---|---|
| Publishes a policy PDF | Publishes evidence — audit scores, incidents, and response times — updated over time. |
| Names a Responsible AI lead | Names accountable owners per system, with override and shutdown authority documented. |
| Runs an internal review | Invites independent reviewers and publishes their findings, including what was not fixed. |
| Offers a support email | Provides a public reporting channel with a triage SLA and a visible resolution status. |
| Reports once a year | Maintains a living Transparency Center that anyone — press, regulators, affected people — can inspect any day. |
Adopt the standard
Bring your organization's AI systems onto the Ethos Registry and publish structured trust indicators.
trust@ethosairegistry.org →Commission a review
Engage an independent review of a high-stakes AI system and publish the findings.
reviews@ethosairegistry.org →Procurement partnership
Use Ethos trust indicators in your RFPs, vendor questionnaires, and evaluation rubrics.
procurement@ethosairegistry.org →