Colorado AG Oct 6 redline of ADMT and Chatbot Safety Act rules adds a 'materially influence' presumption and defines human reviewers by override authority

Colorado Department of Law (Attorney General) · State regulatory oversight of automated decision-making · Colorado

Publisher: Orrick InfoBytes
Original publication: 2026-10-09
Source retrieved: 2026-10-10
Event date: 2026-10-06

What happened

On Oct 6, the Colorado AG released an interim redline of proposed rules for the ADMT Act (SB 26-189) and Chatbot Safety Act (HB 26-1263), ahead of an Oct 26 hearing; comments accepted through the hearing; laws take effect Jan 1, 2027. The draft adds a presumption that an output 'materially influences' a consequential decision when the decision-maker reviews it and it matches the outcome, rebuttable as a 'de minimis factor'. It drops the requirement to state principal reasons for adverse outcomes, replacing it with a specific description of the technology's role, removes the stay of adverse outcomes pending correction or human review, and lets human reviewers be internal or external if they have actual authority to override.

Why it matters

A state AG acting as AI regulator; the human-review and adverse-outcome definitions are transferable benchmarks for state agencies writing rights-affecting human-in-the-loop rules (e.g., Utah EO 2026-08). Whether Colorado agencies themselves are covered deployers is not addressed in the source.

Evidence and measured results

Opened the Orrick InfoBytes summary (Oct 9) via curl; the AG redline itself was not opened this run.

Limitations and uncertainty

Law-firm secondary summary of an interim draft; rules not final and may change after Oct 26 hearing. Primary redline not inspected. Applicability to state agencies unclear; primarily private-sector deployer obligations.

Put this evidence to work

Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-09; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.

Sales

Role takeaway

Customer problem: Agencies need a defensible definition of meaningful human review for AI-influenced decisions.
Stakeholders: Agency general counsel
Program directors for benefits/licensing decisions
State CIO/AI office
Attorney general staff
Discovery questions: Which decisions use AI scores or recommendations?
Do reviewers have documented override authority?
Can systems describe the technology's role in an adverse outcome?
Are agencies covered deployers under state law?
Value hypothesis: Mapping workflows to an override-authority standard reduces ambiguity in rights-affecting AI use.
Potential engagement: Workflow and reviewer-authority assessment for a few high-impact decision processes.
Unsupported claims: That Colorado agencies are covered by these rules
Final rule content
Reduced legal risk from adopting the draft standard

Pre-sales engineering

Role takeaway

Fit: Relevant to decision-support systems producing scores or recommendations.
Architecture and integration: Capture output, reviewer identity, authority and final decision in audit logs; template role-of-technology disclosures.
Prerequisites: Decision-workflow inventory
Reviewer delegation records
Primary redline review
Constraints: Interim draft subject to change
Colorado-specific law
Agency coverage unclear
Security: Protect audit logs containing personal data and scores; restrict access to reviewers and auditors.
Proof of value: Sample decision trace showing reviewer override capability
Generated disclosure matched to a test adverse outcome
Classification of each tool as decision-influencing or summarization-only

Delivery

Role takeaway

Work: Inventory AI-influenced decisions, document reviewer authority, stand up disclosure handling.
Dependencies: Final rules after Oct 26 hearing
Counsel interpretation of coverage
Program staffing
Ownership: Agency counsel with program and AI governance leads.
Skills and adoption: Reviewers trained on override authority and documentation.
Governance checkpoints: Coverage determination
Reviewer delegation sign-off
Pre-effective-date readiness review
Acceptance criteria: Proposed: 100% of in-scope workflows list a reviewer with override authority
Proposed: information requests answered within 15 days
Proposed: disclosures describe the technology's role for each adverse outcome
Risks: Final rules differ from redline
Rubber-stamp review despite authority
Over-reliance on secondary summary

Implementation considerations

Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.

Architecture and integration

What must connect, and where does the AI sit in the workflow?

Systems producing scores, rankings or recommendations reviewed by decision-makers fall under the presumption; summarization-only tools are treated differently.

Governance

Who approves, reviews and stays accountable for outcomes?

Defines meaningful human review by override authority rather than review existence, and loosens disclosure from reasons to role description.

Security and privacy

What data, permissions and controls need testing?

Correction rights narrowed to instructions for correcting personal data, excluding scores and predictions.

Accessibility and workforce

Who is affected, and what skills or accommodations follow?

Human reviewers must hold real override authority, a staffing and delegation question.

Procurement

What should contracts, pricing and exit terms secure?

Agencies buying decision-support tools may need vendors to support role-of-technology disclosures.

Operating model

Which teams own the service once it runs?

15-day response window for information requests implies a staffed disclosure process.

Publication history

2026-10-09State Government · Issue 342 resources

Read preserved resource versions (JSON)

Stable resource ID: colorado-ag-admt-rules-redline-human-review-authority-20261006