Requirements Checklist

Original SB 24-205 vs. Revised Working Group Framework

Impact Assessment

Modified

Original Law

Developers and deployers must complete impact assessments for all high-risk AI systems before deployment

Revised Framework

Risk-tiered approach: full assessment for high-risk only; streamlined assessment for medium-risk; self-certification for low-risk

Agency Action Required

Categorize all AI systems by risk tier; complete appropriate assessment level

Transparency & Notice

Enhanced

Original Law

Consumers must be notified when interacting with AI and when AI makes consequential decisions

Revised Framework

Maintains consumer notification requirements; adds specific disclosure templates for government use

Agency Action Required

Implement disclosure notices on all citizen-facing AI systems

Human Review

Modified

Original Law

Meaningful human review required for all consequential AI decisions

Revised Framework

Risk-proportionate human oversight; high-risk requires human-in-the-loop; medium-risk allows human-on-the-loop

Agency Action Required

Establish review protocols matched to risk level of each system

Data Correction

Enhanced

Original Law

Individuals must be able to correct data used by AI systems that affect them

Revised Framework

Maintains correction rights; adds 30-day response requirement and appeal process

Agency Action Required

Create data correction request workflows for each high-risk system

Bias Testing

Enhanced

Original Law

Regular testing for algorithmic discrimination required

Revised Framework

Annual bias audits for high-risk systems; defines protected classes and testing methodology

Agency Action Required

Schedule annual audits; document testing methodology and results

Incident Reporting

New

Original Law

Not explicitly required

Revised Framework

New requirement: AI-related incidents must be reported to AG within 72 hours

Agency Action Required

Establish incident response plan and reporting pipeline

Vendor Obligations

Enhanced

Original Law

AI developers must provide documentation to deployers

Revised Framework

Enhanced: developers must provide model cards, testing results, and known limitations; state contracts must include AI disclosure clauses

Agency Action Required

Update procurement contracts to include AI disclosure requirements

Enforcement

Modified

Original Law

AG exclusive enforcement; private right of action for violations

Revised Framework

AG enforcement with safe harbor for good-faith compliance efforts; no private right of action during 18-month transition

Agency Action Required

Document good-faith compliance efforts for safe harbor eligibility