No US state or Canadian province has enacted an AI-specific rule of professional conduct. Instead, regulators on both sides of the border have issued guidance mapping generative AI onto the existing rules — most comprehensively the ABA's Formal Opinion 512 (July 2024) — and the consistent message is that competence, confidentiality, supervision, candor, and fee rules govern AI exactly as they governed email, cloud storage, and outsourcing before it. The landscape is evolving quarter by quarter: state bars and law societies keep issuing opinions, practice resources, and task-force reports, so everything below is a map, not a substitute for checking your own regulator's current position.
This article is the regulatory backbone of our AI in Legal Practice library — the practical workflows sit in companion pieces like Can Lawyers Use ChatGPT?.
Competence — Rule 1.1. Comment 8, adopted in some form by 40 states, requires lawyers to keep abreast of "the benefits and risks associated with relevant technology." Formal Opinion 512 reads this to mean a lawyer using generative AI must understand its capabilities and limitations — including hallucination — at a level sufficient to use it responsibly, or must acquire that understanding, associate with someone who has it, or not use the tool. Competence does not require becoming an engineer; it requires knowing enough not to file unverified output.
Confidentiality — Rule 1.6. Lawyers must make reasonable efforts to prevent unauthorized disclosure of information relating to a representation. Op. 512's key move: inputting client information into a self-learning tool that may retain or train on it can require the client's informed consent, and general engagement-letter boilerplate will not carry that weight for consumer tools. Enterprise deployments with contractual no-training and retention terms change the analysis.
Supervision — Rules 5.1 and 5.3. Managerial and supervisory lawyers must ensure the conduct of subordinate lawyers and non-lawyer assistance is compatible with professional obligations — and regulators treat AI tools as falling within the Rule 5.3 framework. That translates into firm-level duties: written AI policies, approved-tool lists, training, and review of AI-assisted work product. A partner cannot delegate to an algorithm what she could not delegate to an unsupervised assistant.
Fees — Rule 1.5. Fees must be reasonable, and Op. 512 confirms two concrete rules: a lawyer may not bill hourly for time not actually spent because AI compressed the task, and AI costs may be passed through only as disclosed, actual expenses. Candor to tribunals (Rule 3.3) and meritorious-claims obligations (Rule 3.1, plus FRCP 11) complete the picture for court work.
Two further Model Rule threads complete the US picture. Candor to the tribunal (Rule 3.3) and the duty to assert only meritorious claims (Rule 3.1) are the rules the sanctions cases actually invoke — a fabricated citation is a false statement to a court however it was generated. And the communication duty (Rule 1.4) is where several state opinions locate the client-disclosure question: whether a client would reasonably want to know that generative AI played a significant role in the matter is analyzed as part of keeping the client reasonably informed, which is why the emerging engagement-letter practice discussed below is spreading even in jurisdictions that mandate nothing explicitly.
State-level activity layers on top: California's COPRAC practical guidance (November 2023) was first; Florida's Ethics Opinion 24-1, New York's bar task-force report, Texas Opinion 705, and a dozen more have followed, all broadly consistent with the ABA framework but differing on details like consent and disclosure — which is exactly why the check-your-jurisdiction habit matters.
Canada's structure parallels the American one. The Federation of Law Societies' Model Code supplies the substance: Rule 3.1-2 (competence, with commentary on technological competence), Rule 3.3-1 (confidentiality), Rule 6.1-1 (direct supervision of delegated work), and Rule 3.6-1 (fees must be fair and reasonable and disclosed in a timely fashion). Provincial law societies have translated these into AI-specific guidance:
Common threads across the Canadian guidance are worth internalizing because they preview where regulation is heading in both countries: verify all output before relying on it; treat consumer AI tools as third-party disclosures for confidentiality purposes; supervise staff and student use explicitly; consider whether the client should be told; and keep records of what tools were used for what. The LSO materials also flag two issues US guidance touches more lightly — bias in AI output as a competence and professionalism concern, and the access-to-justice implications of AI-assisted services — signalling that Canadian regulators are thinking about AI as a practice-transformation issue, not merely a risk-management one.
The Canadian emphasis differs slightly in flavor — more weight on court practice directions and on the public-protection framing — but the operative duties are the same five as in the US, and a firm policy written to the stricter reading of either country's guidance will generally satisfy both.
Three disclosure trends are worth tracking. First, court disclosure: certification and declaration requirements for AI-assisted filings are proliferating through standing orders and practice directions rather than rules of court. Second, client disclosure: regulators increasingly suggest — and Op. 512 requires in defined situations — telling clients when generative AI plays a significant role in the representation, particularly where confidentiality or fees are affected; forward-looking firms are adding a technology clause to engagement letters now. Third, regulator reporting: several law societies have signaled interest in how firms use AI, and insurer questionnaires (LawPRO and US malpractice carriers alike) now routinely ask about AI policies.
Malpractice insurers are becoming the de facto enforcement layer faster than regulators. LawPRO in Ontario and major US carriers now publish AI risk-management guidance, ask about firm AI policies on applications, and — in some markets — price or condition coverage on the answers. A firm with no written policy is increasingly conspicuous, and in a future claim arising from AI-assisted work, the absence of a policy will read as the absence of the "reasonable efforts" every confidentiality and supervision rule requires. The cheapest insurance available is a one-page policy actually followed.
The practical to-do list: adopt a written AI policy mapped to the five duty clusters above; choose enterprise-tier tools with no-training terms; mandate citation verification; add engagement-letter language; log AI-assisted filings; and calendar a semi-annual review of your own regulator's guidance, because this survey will be incomplete within months of any given reading. Firms should treat AI governance the way they treat trust accounting — boring, documented, and audited. And because the same regulators' members are also being found through AI, pair internal governance with external visibility: our ChatGPT for law firms hub covers that side, the free AI Visibility Checker benchmarks it, and LexScale.ai builds the whole system.
LexScale.ai helps law firms across Canada and the United States adopt AI for growth — from client-facing intake and content systems to the visibility that puts your firm inside AI answers.
Book a Free Strategy Call →This article is general information, not legal or ethics advice. Professional-conduct rules on AI are evolving and vary by jurisdiction — always verify current requirements with your state bar, law society, or regulator before adopting any AI workflow.
Related Articles
Can Lawyers Use ChatGPT? Ethics, Risks & Workflow · Client Confidentiality and AI Tools for Lawyers · Court Rules on AI in Filings: US & Canada Guide · Billing Ethics for AI-Assisted Legal Work · AI Contract Drafting for Lawyers: What Works · AI Document Review & Discovery: From TAR to LLMs
Ready to grow your firm with AI?