Compare

Not every certification
is the same certification.

Buyers, regulators, and the public face four broadly available ways to evaluate whether an AI system is trustworthy. Only one of them is structurally designed to be both independent and publicly verifiable.

CriterionEthicalityConsulting-firm auditsSelf-attestationRegulator-only compliance
Published, public standard

AIMSS in full

Proprietary methodology

Vendor whitepaper

Statute only

Public registry of certified organisations

Where regulator publishes

Independence policy (no consulting to audited clients)

Same firm sells both

Unannounced surveillance visits

Where authority empowered

Public complaints register
Suspensions & withdrawals published
Transparent, scoped fees

Quoted in engagement letter

Bespoke engagements

Free

Statutory fees

Buyer-side certification for AI users

Ethical Use track

Adopt-into-standard process for public proposals

Policy Institute

Rule-making

Independent appeals panel

Judicial review

Assessor rotation rules

✓ Present · ◐ Partial · — Absent. No firms or individuals are named.

What this means

The questions to ask any AI audit.

Is the standard public? If you cannot read it, you cannot challenge it.

Can I look up the certificate? If there is no registry, claims are unverifiable.

Who pays the assessor, and who else do they sell to the same client? If the assessment firm also consults them, the report's independence is structural fiction.

What happens between audits? A certificate that is only checked every three years is a snapshot. Continuous integrity is the standard the public deserves.

Where do complaints go? If there is no public channel, problems stay private until they explode.

Read further

See how the structure works in practice.