An open standard for AI competence
The AI Competence Framework defines what people need to know and be able to do to work effectively with AI. It is organised by domain, proficiency level and professional role, and is published in full.
The framework is free to use, adapt, translate and build on. AI Certification Standards issues no credentials and assesses no candidates. Training providers, employers and awarding bodies apply the framework independently.
Version 1.29 · issued 23.08.2026 · open for comment until 30.11.2026 · licensed CC BY 4.0
Areas of competence
From AI foundations and data handling through to evaluation, governance and security.
Aware to Advanced
Each level is defined by autonomy, complexity, judgement, scope and accountability — not by years of experience.
Professional context
Additional statements for those who instruct, enable or audit, applied on top of the core domains.
Find a competence statement
Every statement in the framework has a permanent identifier and a permanent address. Statements are never renumbered and never deleted.
Statements published 255 · across 8 of 8 domains, three overlays and five extensions
This is the statement index, part of the specification. The register of conformance claims is a separate record and holds no entries until version 1.29.
Map a course
Publish a coverage map stating which statements your course teaches and which it assesses.
Mapping a courseDefine a role
Write requirements against defined levels rather than against tool names.
Guidance for employersBuild a credential
Construct a credential specification on a shared, public and versioned reference.
The specification standardCheck a claim
Establish what a proficiency level requires before paying for a course that claims it.
Reading a coverage mapCurrent state of the framework
All eight domains are entered in the statement index at version 1.29. Every statement carries a permanent identifier and is never renumbered.
| Ref | Domain | Statements | Status |
|---|---|---|---|
| D1 | AI foundations | 22 | PUBLISHED |
| D2 | Data and information | 23 | PUBLISHED |
| D3 | Interaction and elicitation | 19 | PUBLISHED |
| D4 | Automation and integration | 22 | PUBLISHED |
| D5 | Solution design | 22 | PUBLISHED |
| D6 | Evaluation and assurance | 22 | PUBLISHED |
| D7 | Ethics, governance and professional practice | 22 | PUBLISHED |
| D8 | Security | 25 | PUBLISHED |
Four proficiency levels are defined: Aware, Applied, Proficient and Advanced. A fifth level is reserved and deliberately undefined.
Notices
Releases, consultations, conformance decisions and recusals, dated and attributed to the office that issued them. Nothing is removed from the record.
Articles
Explanation and reasoning, bylined and revisable. No article is part of the framework, and this stream is never merged with the record beside it.
Who publishes this
AI Certification Standards is a nonprofit standards body. Its board, funding and conflict-of-interest policy are published in full, including the relationships between this organisation and the certification and training bodies founded alongside it.
Where a decision involves a related party, the conflicted parties are recused and the recusal is recorded.