UAIF
Unified AI Incident Framework
Incident interchange, classification and severity framework with a seven-layer architecture, conformance profiles and machine-readable schema.
Purpose and scope
Incident interchange, classification and severity framework with a seven-layer architecture, conformance profiles and machine-readable schema.
Use the domain index below to navigate to domain-level summaries and individual field, modifier, profile and test pages. Each detail page includes the available objective, UAIF requirement, validation approach, evidence expectations, mapping information and related records.
Architecture sections and artifacts
Incident Identity & Workflow
Deduplication, audit trail, confidence scoring
L1Causal Separation Chain
Isolates origin from manifestation
L2Severity Scoring
Impact-dominant severity with documented limitations
L3Acute vs. Chronic Harm
Immediate impact + long-term systemic degradation
L4Incident vs. Vulnerability
Static exposures vs. dynamic occurrences
L5Generative, Agentic & Control Plane
RAG leakage, hallucinations, agent escalation, MCP, OAuth, policy enforcement
L6Regulatory & Cross-Sector Metadata
Jurisdictional routing, sector classification
SCSeverity & Harm Scoring
Impact-dominant severity calculation with provisional harm weights and transparent limitations.
CMContext Modifiers
Attack-vector-specific modifiers applied through capped diminishing contribution logic.
CPConformance Profiles
Profile-specific field obligations for Core, Enterprise, Regulatory, SOC/SIEM, and AI Security Extension uses.
CTConformance Testing
Minimum validation requirements for a conformant UAIF implementation.
Using this publication
- Use the left navigation for the complete framework tree.
- Search by artifact ID, domain, evidence type or technical term.
- Use control-level links as stable citation targets.
- Consult the publication status notice before relying on the material operationally.
Launch announcement
Read the GAISSF Ecosystem launch announcement and Founder Note.