AI CORE / DOCUMENTATION
A shared language for verifiable AI systems
AI CORE provides a stable vocabulary, assessment profiles and an evidence record format. It helps prepare an assessment that makes clear what was examined, against which criteria and with what evidence.
Three parts of one reference
Ontology 2.0 names the audited object, plan, executor and evidence. Profiles 1.0.0 select specific control objectives and describe their application. Record schema 1.0.0 specifies how to communicate a result without confusing a claim with evidence. These parts have separate versions: changing the website does not change the meaning of the ontology.
Who this documentation is for
Operators can find guidance on preparing evidence. Assessors can identify the scope, conditions of use and limitations of profiles. Developers and software agents can retrieve the same requirements as JSON, select an exact version and use it without depending on the availability of this website.
How to start
- Select a profile based on system capabilities, not its AI marketing label.
- Record the assessment scope, system snapshot, permissions and exclusions.
- For each applicable requirement, identify a method and the expected evidence.
- Obtain evidence and distinguish self-reported claims from observations and independent conclusions.
- Publish only a safe subset. Explicitly identify missing or insufficient evidence.
What a stable release means
A stable profile is a completed reference specification with identifiers, criteria and interpretation rules. Publishing it does not constitute an audit of AI INDUSTRY or confirmation of another system's security. A specific assessment produces its own separate result.
AI CORE does not operate an audit program, accept private evidence packages or issue certificates. It provides public references; execution methods and professional conclusions belong to the assessor.