Colorado’s New AI Act Targets Automated Decision-Making for Consequential Decisions

Why It Matters

The legislation grants employees and job applicants enforceable AI‑related rights, imposing new disclosure duties on employers and limiting contract‑based liability shields, thereby reshaping AI governance in the workplace.

Key Takeaways

• SB 26‑189 defines “covered ADMT” across seven consequential domains
• Pre‑use notice and 30‑day adverse‑outcome disclosure replace risk‑assessment duties
• Employees and applicants now receive AI‑related rights under the act
• Indemnity clauses shielding developers from discrimination liability are void
• Enforcement includes a 60‑day cure period and no private right of action

Pulse Analysis

Colorado’s AI regulatory journey has been turbulent. The 2024 AI Act, signed after a contentious stakeholder review, imposed extensive duties such as risk‑management programs, impact assessments, and algorithmic discrimination notices. Persistent legislative deadlock and a governor‑led working group prompted a rewrite, culminating in SB 26‑189. By focusing on "covered ADMT"—AI tools that materially influence decisions in education, employment, housing, finance, insurance, healthcare, and essential government services—the state narrows its reach while still targeting the most consequential uses of the technology.

The new compliance framework pivots from prescriptive processes to transparency. Developers must supply documentation on intended uses, data sources, and limitations, while deployers provide a clear pre‑use notice at the point of interaction and a plain‑language disclosure within 30 days of any adverse outcome. A five‑part "meaningful human review" standard ensures that a qualified individual can override AI recommendations, access relevant evidence, and receive sufficient system context without exposing trade secrets. Liability is no longer jointly assigned; fault follows the party responsible for the specific use, and any contract term that indemnifies a developer or deployer against antidiscrimination claims is void, forcing companies to renegotiate vendor agreements.

For employers, the act delivers both clarity and new obligations. Inventorying AI tools to determine covered status, revising contracts to remove prohibited indemnities, and bolstering human‑review processes are immediate priorities before the Jan 1, 2027 deadline. Although the law eliminates mandatory bias audits, discrimination risk under Title VII, the Colorado Anti‑Discrimination Act, and emerging state AI statutes remains. Federal preemption efforts, spurred by a 2025 DOJ AI Litigation Task Force and a White House AI policy push, could later reshape state regimes, making Colorado’s approach a bellwether for how employers navigate evolving AI governance across jurisdictions.