The Colorado AI Act (SB24-205) was the first state AI law in the US. In May 2026 SB26-189 repealed and re-enacted those rules - the new automated decision-making regime takes effect on 1 January 2027.
- Colorado passed the first state AI law in the US (SB24-205, 2024): 'reasonable care' against algorithmic discrimination.
- In May 2026 SB26-189 repealed and re-enacted the rules; the new regime takes effect on 1 January 2027.
- The framework: impact assessments, notice, human review on appeal; enforced by the attorney general.
Colorado went first. The first US state with its own AI law (SB24-205, passed 2024) - 'reasonable care' from developers and deployers of high-risk systems against algorithmic discrimination in 'consequential' (decisions with real impact on a person) decisions. The original start date for the duties was set for 1 February 2026.
Updated 1 July 2026: in May 2026 the state passed SB26-189, which repeals and re-enacts those rules as a new law on 'automated decision-making' (ADMT). The new regime takes effect on 1 January 2027, and the attorney general must adopt rules before that date. The old 2026 deadlines are gone - the real date is now 2027. Our first version (with a 1 February 2026 start) is outdated. We're fixing it here, in the corrections log.
The important detail is the focus - not 'dangerous AI' in general, but discrimination: in employment, credit, housing. The law doesn't ask whether the model is smart. It asks whether it treats people fairly. That's the practical question.
While the EU builds the big general framework, one American state is betting on something else - a narrow law, built from a specific harm. Different paths, but both put rules where until recently there were none - on decisions that actually affect people.