FRAMEWORK SUITE
Framework comparison
The frameworks are connected but not interchangeable.
Comparison table
| Framework | Full name | Scale | Primary role | Publication status |
|---|---|---|---|---|
| GAISSF | Global AI Safety and Security Framework | 9 domains | Implementation-independent, evidence-based governance and assurance framework covering technical, organizational, human, regulatory and physical-AI risk. | Final Publication v1.0 |
| UAIF | Unified AI Incident Framework | 11 sections | Incident interchange, classification and severity framework with a seven-layer architecture, conformance profiles and machine-readable schema. | Final Publication v1.0 |
| AI-IRF | AI Incident Response Framework | 19 domains | Operational incident response architecture spanning preparation, detection, classification, containment, recovery, notification and learning. | Final Publication v1.0 |
| PAI-SF | Physical AI Security Framework | 12 domains · 41 controls | Security and assurance framework for AI systems whose decisions can produce physical effects, covering sensing, control, actuation, safety interfaces, communications, autonomy and recovery. | Final Publication v1.0 |
Selection guidance
Use GAISSF when establishing governance, controls and assurance. Use UAIF when recording and exchanging incident information. Use AI-IRF when designing and executing operational response capability. Use PAI-SF when AI decisions can cause or materially influence physical-world actions, movement, actuation, machinery or safety-relevant effects.