Colorado AI Ethics Rules for Lawyers

The official instruments

Rule text, not commentary about rules

Most states have answered generative AI with an ethics opinion, a guidance document, or a committee report. On January 8, 2026, the Colorado Supreme Court answered with an amendment: Rule Change 2026(02), adopted en banc and effective immediately, writes technology, expressly including artificial intelligence, into the Colorado Rules of Professional Conduct. That distinction is the headline. An opinion interprets rules that never mention AI; Colorado’s rulebook now mentions it. In this dataset that makes Rule Change 2026(02) a first, and the tracker phrases the first carefully: it is the first AI-specific conduct-rule amendment adopted by a state supreme court that we have recorded, verified against the order’s primary PDF on the judicial branch’s site, not a claim about every survey ever compiled. California’s proposed amendments, by comparison, remained proposals at our last verification, which makes the two states the cleanest contrast in the country between pending and done, and Arkansas sits in the same pending column.

Paragraph 20A: the responsibility clause

The amendment’s centerpiece is a new Preamble and Scope paragraph 20A, and its rule is blunt: technology, including artificial intelligence, does not diminish a lawyer’s responsibilities, and a lawyer who uses technology directly or indirectly in delivering legal services may be disciplined for violations that result. Two drafting choices deserve attention. First, the paragraph regulates technology generally and names AI as an instance, so it will not age out when the next tool category arrives. Second, the direct-or-indirect language follows the work rather than the interface: no lawyer steps outside the rule by putting a vendor, a platform, or an assistant between themselves and the model. Discipline for resulting violations is the operative consequence, and it converts every AI deployment decision inside a Colorado practice into a conduct-rule decision. The supervision analysis covers the adjacent duties that the indirect-use language most obviously engages.

Comment 9’s map of the rulebook

Alongside the Scope amendment, Rule Change 2026(02) reworks the competence commentary. Rule 1.1’s comment 8 on technological competence is amended, and a new comment 9 does something every Colorado firm should exploit: it catalogs the rules AI use can implicate, naming Rules 1.4 (communication), 1.5 (fees), 1.6 (confidentiality), 3.1 (meritorious claims), 3.3 (candor), 5.1 through 5.3 (supervision), 7.1 (communications about services), and 8.4(g) (discriminatory conduct). A compliance review can run straight down that list, which is presumably the point of publishing it. Comment 9 also confirms that reliance on technology does not diminish the lawyer’s independent judgment, the same principle the hallucination sanctions guide traces through the caselaw: model output is raw material for judgment, never a substitute. For the two list entries firms ask about most, the confidentiality breakdown and the billing ethics analysis map directly onto comment 9’s own citations.

A committee already drafting what comes next

Rule Change 2026(02) is not positioned as Colorado’s last word, because the state built a pipeline for more. On September 15, 2025, Chief Justice Márquez launched the Legal Technology Advisory Committee, announced by the judicial branch’s own press release, with initial recommendations due to the court by October 1, 2026. A standing committee with a due date is worth taking literally: Colorado expects further movement, and the next deliverable lands within months of this page’s last verification. For secondary context, the Colorado Bar Association’s Colorado Lawyer has published a feature-length analysis of the 2026 amendments; it is helpful practitioner reading so long as the order itself remains the citation. The near-term work for a Colorado firm is concrete: read the primary order, walk comment 9’s rule list against current workflows, and record the results in a written policy. The policy template structures that exercise, the state comparison table shows how unusual binding text still is, and the changelog will pick up whatever the advisory committee produces once it passes primary-source verification.