Colorado’s New AI Act Targets Automated Decision-Making for Consequential Decisions
Colorado’s New AI Act Targets Automated Decision-Making for Consequential Decisions
Publish Date: 2026-05-26 17:54:00
Source Domain: ogletree.com
- On May 14, 2026, Colorado Governor Polis signed a new AI regulation, SB 26-189, which replaces and repeals the 2024 Colorado AI Act.
- The new law reduces the oversight and compliance requirements from the previous act by simplifying the responsibilities to pre-use notices, post-adverse-outcome disclosures, and a limited number of consumer rights.
- The “consumer” category includes employees and Colorado resident job applicants, extending further protections than the Colorado Privacy Act provided.
- Contracts indemnifying developers or deployers against antidiscrimination violations are void as against public policy.
- The new law will take effect on January 1, 2027.
- SB 26-189 narrows the scope to exclude low-stakes uses and routine decisions of AI, focusing on consequential decisions in employment, housing, education, financial, health, and essential government services.
- “Covered ADMT” refers to automated decision-making technology used to materially influence crucial decisions, excluding simple tools like spell-check.
- Employers should pay attention to the definition of “consumer” and prepare for new notices and rights after adverse outcomes.
- The attorney general holds exclusive enforcement power and has a sixty-day cure window to avoid penalties.
- Private rights of action are not created, and existing antidiscrimination laws remain in force.
- Sectoral exemptions apply to insurers and HIPAA-covered entities, but not for workforce decisions related to employment.
- For employers, the law should provide more clarity by eliminating impact assessment and risk management requirements, with effective date March 9, 2027.
- Employers are encouraged to review AI tools, contracts, and compliance processes in preparation for the new law.
- Federal developments, including potential preemption, are worth monitoring, as they may impact the practical reach of state-specific AI laws.
