AI SECURITY CASES × FRAMEWORKS
Google SAIF 2.0 · Categories and AI Security Cases
Explore AI security cases, sources, and defensive analysis across 14 populated Google SAIF 2.0 categories.
A comprehensive framework identifying 15 key risks to AI systems across the development and deployment lifecycle, published by Google.
- DP · Data Poisoning 5 cases
- UTD · Unauthorized Training Data 3 cases
- MST · Model Source Tampering 11 cases
- EDH · Excessive Data Handling 1 cases
- MXF · Model Exfiltration 6 cases
- MDT · Model Deployment Tampering 5 cases
- DMS · Denial of ML Service 1 cases
- MRE · Model Reverse Engineering 6 cases
- IIC · Insecure Integrated Component 74 cases
- PIJ · Prompt Injection 85 cases
- MEV · Model Evasion 16 cases
- SDD · Sensitive Data Disclosure 80 cases
- ISD · Inferred Sensitive Data 0 cases
- IMO · Insecure Model Output 31 cases
- RA · Rogue Actions 62 cases
An empty category does not establish safety. A category relation does not prove a defense is deployed or effective.