Connecticut AI Ethics Rules for Lawyers

No AI-specific instrument yet

This jurisdiction had issued no AI-specific ethics opinion, court rule, or formal guidance as of our last review. That does not make AI use ungoverned here: the existing rules of professional conduct — competence, confidentiality, supervision, and candor — apply to generative AI the way they apply to any other tool, and ABA Formal Opinion 512 is the reference point most bars borrow in the meantime. The breakdown below covers what that means in practice; the changelog will record the week anything official issues.

The whole index, read to the bottom

A negative claim is only worth publishing if the index behind it was actually read, so on July 19, 2026 we inspected the Connecticut Bar Association’s informal ethics opinions index in full. It lists opinions from 2011 through 2025, and the most recent entry is 25-04, on the duty to report professional misconduct. No opinion in the listing addresses artificial intelligence. The closest the index comes is a pair of pre-generative-AI technology items, 13-07 on cloud computing and 11-03 on storing client information online, which means the CBA’s published technology line last moved roughly a decade before the tools this tracker covers existed in their current form. The index page itself notes that the opinions are advisory, “not authoritative and are not binding on the Statewide Grievance Committee or the courts,” a caveat worth remembering about any CBA opinion that eventually does address AI. The tracker records Connecticut as a dated, verified negative, the posture the methodology page explains: a negative is only as good as the day it was checked, so this one carries its date and a re-verification schedule instead of a claim of permanence.

The Branch wrote policy for its own people

Connecticut’s most concrete AI document comes from the Judicial Branch, and it points inward. The Judicial Branch Artificial Intelligence Committee’s report of July 2024, by the Branch’s own description, delineates policy for the use of AI by Judicial Branch employees: it replaces section 9 of Administrative Policy 1013 with a Responsible AI Use Framework resting on three principles, purposeful use, meaningful guardrails, and workplace empowerment and education. We describe the report the way the Branch itself frames it, and that framing is the point of this section: this is workplace policy for the court system’s own workforce, not guidance addressed to the practicing bar, so it enters this page as judiciary activity and stays out of the instrument dataset. The line is the same one drawn for Delaware, where a bench-and-staff AI policy is described prominently but never counted as attorney guidance, because counting it would misstate what a lawyer in the state can actually be held to. Connecticut practitioners should still know the report exists: it shows the court system building its own AI governance before anyone built any for the bar.

A committee named for the whole legal system

The Judicial Branch also maintains the Committee on Artificial Intelligence in the Connecticut Legal System, the CAIC. The name is broader than the employee policy above, and that breadth is the signal worth tracking: a standing committee scoped to the legal system as a whole is the kind of body from which attorney-facing rules or guidance could eventually emerge. For now it is committee activity, and this page characterizes it only as that: we record its existence and characterize no output, because as of July 19, 2026 no attorney-facing AI instrument from any Connecticut source we checked had issued. If the CAIC’s work produces one, from the courts or through the bar, this page converts from a verified negative into a deep-dive, and the changelog will say so the week a primary source verifies it.

Practicing in Connecticut on the baseline

An empty instrument column changes nothing about the duties. Competence requires understanding what a generative tool does before it touches client work, and the tool risk tiers give that analysis a structure. Confidentiality requires knowing where a prompt goes and what the vendor retains, analyzed before first use, as the confidentiality guide details, and it is no accident that the CBA’s closest existing opinions, cloud computing and online storage, are ancestors of exactly that question. Verification stands between AI drafting and anything filed or sent, the discipline the sanctions caselaw exists to motivate. Billing follows actual time, per the billing analysis, and office-wide oversight runs through the supervision rules. ABA Formal Opinion 512 assembles the framework nationally, the policy template turns it into an internal document, and because individual Connecticut judges may impose chambers-level requirements at any time, the court orders guide covers the per-filing check that no statewide silence excuses.