Kentucky AI Ethics Rules for Lawyers

The official instruments

A numbered opinion that asks the duty-to-use question

Kentucky put its AI position on the record early and formally. KBA Ethics Opinion E-457, adopted by the Kentucky Bar Association Board of Governors on March 15, 2024, is a numbered formal opinion, not a committee article, a set of FAQs, or informal guidance. That distinction matters for citation weight: where several states regulate AI through education or unnumbered website updates, Kentucky answered with a formal instrument that a Kentucky lawyer, a disciplinary authority, or an out-of-state firm can cite by number. The opinion is organized as seven questions, a format that makes it unusually easy to map against a firm’s own checklist, because each holding answers a discrete practical problem rather than narrating a general philosophy.

The opinion’s most distinctive move is on competence. E-457 does not merely say a lawyer must understand AI before using it; it frames the analysis so that declining to use an available tool may itself implicate the duty of competence. That inverts the assumption built into much early commentary, which treated AI as an optional efficiency a careful lawyer could simply avoid. Kentucky’s framing puts the technology on the same footing as other research methods a competent practitioner is expected to be conversant with. The practical reading is not that every Kentucky lawyer must buy an AI subscription tomorrow; it is that ignorance of the category is no longer a safe default, and competence is assessed in light of what the tools can reasonably do.

Disclosure, fees, and the conditions Kentucky attaches

E-457 takes a measured position on disclosure. Routine AI-assisted research need not be disclosed to the client or the court, but that baseline is conditional. Disclosure is owed when the work is outsourced, when the client is being charged for the AI work specifically, or when a court rule requires it. This is a trigger-based duty rather than a blanket one, and the triggers are worth memorizing because they are the difference between permissible silence and a communication violation. A Kentucky firm that treats AI like an ordinary research tool for internal use is on solid ground; the same firm crosses a line if it bills AI work without saying so, hands a matter to an outside AI-driven vendor without telling the client, or ignores a standing order that asks about AI use.

On fees, E-457 is firm. A lawyer must reduce fees when AI shrinks the time actually spent on a task, which is the same principle the billing analysis tracks across states and which ABA Formal Opinion 512 reaches by the same route: a lawyer bills for time worked, and time the machine saved is not time worked. Kentucky adds a specific procedural requirement for cost pass-through: AI expenses are chargeable to the client only under an advance written agreement. That is a documentation rule, not just a fairness principle. A Kentucky firm cannot decide after the fact to bill a research-platform AI charge to a client; the agreement to do so has to exist in writing before the cost is incurred. The cleanest implementation is engagement-letter language addressing AI costs at the outset, which also satisfies the communication side of the same problem.

Confidentiality, verification, and what Kentucky firms should do now

The remaining holdings track the national consensus but are stated as Kentucky duties. Lawyers must safeguard client confidences, which in practice means evaluating whether a given tool exposes or retains inputs before any client information goes into it. Lawyers must verify AI-assisted court submissions, the anti-hallucination duty that every serious instrument now imposes and that the sanctions caselaw made unavoidable. And lawyers must adopt firm AI policies and training, locating supervision at the firm level rather than leaving it to individual discretion.

For a Kentucky firm, the build is concrete. Maintain an approved-tools register that records each tool’s confidentiality and data-handling terms, so the safeguarding duty is documented rather than assumed. Set engagement-letter AI language that handles both the disclosure triggers and the advance-written-agreement requirement for costs in one place. Enforce a verification workflow that checks every citation in a traditional database before filing. Adopt a written, acknowledged AI policy with training, which is what the opinion asks for under supervision and what the policy template implements element by element. The comparison table places Kentucky’s conditional-disclosure posture against stricter and looser states, and the tracker and changelog will move E-457’s entry the week anything supersedes it.

One layer E-457 does not occupy is the courtroom itself. Kentucky has no statewide court rule on generative AI that this tracker has verified, so the obligations that exist at the filing stage are judge-by-judge standing orders rather than a uniform state command. That makes the per-filing forum check a live requirement: before relying on AI-assisted work in any Kentucky court, confirm the assigned judge’s standing orders, because a disclosure or certification requirement can exist at the chambers level even though the state has issued none. The court orders guide covers that check and the kinds of orders to watch for.