Public explainer

Certification Is Not Accreditation

Released July 17, 2026 · Current

Light-tier public explainer. Noncanonical. Authority effect: None. This piece is a readable derivative of Standards Body Terminology, Version 1.0, which governs if anything here differs from it. It is not an adopted standard and does not certify, accredit, or approve anything.

Certification and accreditation are probably the most frequently swapped words in discussions of AI assurance, and the swap is never harmless, because the two answer different questions about different parties.

Certification is third-party attestation that specified requirements have been fulfilled within a defined certification scheme. It is about a subject, such as a product, system, or organization, and it is only ever meaningful relative to its scheme: the rules, criteria, assessment methods, surveillance, and responsibilities that govern it. Certification does not mean universal safety or government approval, and in Standards Body usage the phrase "certified safe" is prohibited, because certification is always against specified requirements and cannot establish universal safety.

Accreditation is one level up: independent recognition that a conformity-assessment body is competent and impartial to perform specified activities within a defined scope. Accreditation is not about the AI system at all. It is about whether the organization doing the certifying, testing, or inspecting is itself competent to do so, and only within its stated scope. An accredited evaluator without a stated scope is a red flag, not a credential.

So a sentence like "the model was accredited" is a category error, and "certified by an accredited body" is actually two separate claims: that requirements in some scheme were met, and that the body attesting to this was independently recognized as competent for that kind of assessment. Each claim can be true or false on its own, and each can be checked. Who defined the scheme? What is its scope? Who recognized the certifier, and for what?

In frontier AI today, the honest answer to most of those questions is that the schemes, scopes, and recognition arrangements largely do not exist yet. That is precisely why the vocabulary matters now: the words are arriving in public discussion before the institutions have, and using them loosely manufactures assurance that no one has actually provided. Standards Body publishes institutional-design research on what such arrangements could look like, and does not itself certify or accredit anything.

Certification Is Not Accreditation

Read next