Classifying AI Incidents — A Practical Guide to UAIF™
AI incident classification is not simply the act of attaching a label after something goes wrong. An organisation may observe anomalous or harmful system behaviour before it knows the cause, the full consequence, whether the condition is an incident or a vulnerability, or whether a regulatory reporting obligation exists. UAIF™ provides a machine-readable incident interchange, classification, severity-scoring, and conformance architecture for AI incidents. Its seven-layer architecture separates identity and…
Purpose
AI incident classification is not simply the act of attaching a label after something goes wrong. An organisation may observe anomalous or harmful system behaviour before it knows the cause, the full consequence, whether the condition is an incident or a vulnerability, or whether a regulatory reporting obligation exists. UAIF™ provides a machine-readable incident interchange, classification, severity-scoring, and conformance architecture for AI incidents. Its seven-layer architecture separates identity and…
Methodology Note
This is an adoption and practitioner-orientation guide derived from ODA3 Institute's UAIF™ v1.0 technical architecture. It does not replace the controlled UAIF specification or machine-readable schema. Examples are hypothetical and do not represent client incidents, proprietary telemetry, deployment outcomes, or incident-frequency statistics. Where UAIF uses controlled terminology, this guide preserves that terminology. Where the guide adds explanatory workflow language, that language is informative rather than a new UAIF lifecycle or taxonomy.
Source and Traceability This adoption guide is an informative WHP publication. It does not create or modify UAIF normative requirements. ● Controlling technical source: ODA3-2026-06-TCR-STD-002 — UAIF™ v1.0 Technical Specification — Final Publication v1.0. ● Machine-readable companion: TCR-STD-003 — UAIF™ schema specification / schema package. ● Current public status: UAIF™ v1.0 is a Final Publication (TCR-STD-002, TCR-STD-003). Final status means the normative architecture, controlled vocabulary and conformance model are frozen under ODA3 Institute change control — it does not by itself mean every sector calibration, legal interpretation or empirical reliability claim has been validated in production. ● Operational reliance: before implementation, confirm the current schema version, publication state, checksum/release metadata, GEL terms, and any open validation gates in the ODA3 Publication Catalog and JSON Schema Center. COM-012 reaching a final publication state as a practitioner guide is a separate control from UAIF™'s own publication status, and the two happen to align in this revision: both are now final. Document publication status and framework maturity remain separate controls that must not be conflated going forward — a future COM-012 revision could still be FINAL while UAIF™ moves to a later version, or vice versa. Framework maturity is stated in full above and in the Publication Boundary at the end of this guide; where it is referenced again in the body, it is noted briefly rather than restated in full.
HOW TO READ THIS GUIDE All field names, values and examples shown in this guide are illustrative. The normative authority for required, conditional and prohibited fields is the TCR-STD-003 JSON Schema for your organisation's selected conformance profile, together with the current release state published in the ODA3 Publication Catalog and JSON Schema Center. This guide never overrides the schema; where the two differ, the schema governs.
ODA3 Institute | Adoption Guide | Classifying AI Incidents: A Practical Guide to UAIF™
Use this guide when…
Use this guide when you need practitioner-oriented guidance within the stated UAIF scope, while retaining the underlying framework, standard and evidence boundaries.
Intended audience: Practitioners responsible for AI governance, security, implementation, evidence or assurance within the guide’s stated scope.
What this guide supports
Structured practitioner understanding and preparation within its stated scope. It should be read with the relevant normative framework and current ODA3 documentation.
Incident sequence: Recognize → Classify with UAIF → Respond through AI-IRF.
What it does not establish
Use of this guide does not by itself establish implementation completeness, control effectiveness, conformity, certification, independent assurance, regulatory approval, legal compliance, ODA3 approval or authorization to use controlled marks.
Download
Notably Absent
No claim of certification, regulatory approval, legal compliance or independent assurance is created by publication or use of this guide.