The Colorado Attorney General’s Office released a revised draft of its Automated Decision-Making Technology and Chatbot Safety rules today, hitting the September 23 deadline it publicly committed to when it filed the first proposed rules on August 11. Enforcement starts January 1, 2027. That’s 100 days.

SB 26-189, which repeals and reenacts Colorado’s 2024 AI Act, defines ADMT as technology that processes personal data to generate scores, rankings, or recommendations that guide decisions. Per SB 26-189, “consequential decisions” span education, employment, housing, financial or lending services, insurance, health-care services, and essential government services and public benefits. HB 26-1263, the Chatbot Safety Act, layers on age-estimation, AI-disclosure, and crisis-response duties for consumer-facing conversational AI. Violations are deceptive trade practices under the Colorado Consumer Protection Act, enforceable solely by the AG.

The live question, flagged by Seyfarth Shaw, is when ADMT “materially influences” a decision. The August draft floated two competing standards and asked commenters to pick one. Both carry a rebuttable presumption whenever the AI output ranks, scores, or screens data that a human reviewer sees.

That presumption is what pulls a 5-to-30-person shop into scope. An AI lead scorer sorting Colorado inbound, an applicant filter cutting a hiring pile, a support chatbot handling refund requests, each can qualify as covered ADMT touching a consequential decision. Deployers then owe the consumer a plain-language notice within 30 days of any adverse outcome, plus meaningful human review.

Written comments close October 26 at 11:59 p.m. MST, with the rulemaking hearing the same day. It’s the last window to narrow the “materially influences” standard before the text hardens. Monday’s earlier explainer walked through what to watch for. Today’s revision is what the comment period will actually argue over.

Sources

Sources