Risk Assessment
Requirement to assess and document the risks posed by AI systems, including potential harms, bias, and impacts on affected individuals.
What Counts
- Mandatory risk assessments before deployment
- Algorithmic impact assessments
- Documentation of potential harms and mitigations
- Ongoing risk monitoring
- Risk classification and categorization
What Does Not Count
- Generic technology risk assessments without AI-specific criteria
- Self-assessment without documentation requirements
- One-time assessments without ongoing monitoring
Related Terms
- Impact assessment
- Risk management
- Algorithmic audit
- Risk classification
Implementing Regulations
| Regulation | Jurisdiction | Status | Provisions |
|---|---|---|---|
| California CCPA ADMT Regulations | California | enacted | 1 |
| EU AI Act | European Union | phased enforcement | 1 |
| New York RAISE Act | New York | enacted | 1 |
| NIST AI Risk Management Framework | United States | voluntary | 1 |