ABA and State Bar Reference — L2 Technology Solutions
Current as of September 2026. Legal information, not legal advice.
No bar association bans generative AI. Every instrument below applies existing rules of professional conduct to AI rather than creating new rules. ABA Formal Opinion 512 (July 29, 2024) is the closest thing to a national baseline; a growing number of states have layered their own opinions, guidance or court rules on top, and the rest rely on the rules they already had.
We’ve assembled the below guidance to be helpful as you begin to consider a thoughtful implementation of AI. Please reach out to us with feedback, specific questions, and for help brainstorming a plan specific to your firm to prioritize safety and compliance while beginning to tap into the massive time savings and efficiency gains of Generative AI.
On this page
ABA Formal Opinion 512: Generative Artificial Intelligence Tools (ABA Standing Committee on Ethics and Professional Responsibility, July 29, 2024)
The ABA's first and, as of this writing, only formal opinion on generative AI (subsequent Opinions 514–518 cover non-AI topics). Its core message: lawyers may use generative AI, but every duty owed to clients, courts and the profession still applies. The opinion maps six Model Rules onto AI use:
Competence (Rule 1.1) — a reasonable understanding of the tool's capabilities and limits. Expert-level knowledge is not required; blind reliance is not permitted. Includes keeping up as tools change.
Confidentiality (Rule 1.6) — evaluate the tool's terms of use, data retention and whether inputs are used to train the model. Informed client consent is required before entering client information into a self-learning tool; boilerplate engagement-letter language is not enough.
Communication (Rule 1.4) — disclose AI use when it is material to the representation, when the client asks, or when the engagement terms require it. Routine, embedded uses generally do not require disclosure.
Candor and meritorious claims (Rules 3.1, 3.3, 8.4) — independently verify all output, especially citations and quotations, before submitting anything to a tribunal. Comply with any court disclosure orders.
Supervision (Rules 5.1, 5.3) — managerial lawyers must establish firm policies and training on AI use; supervising lawyers are responsible for AI use by lawyers and nonlawyers under them.
Fees (Rule 1.5) — fees must reflect time actually spent. A lawyer may not bill hourly clients for time the tool saved, and may not charge clients for time spent learning a tool. Tool costs may be passed through only at actual cost and with the client's agreement.
Opinion 512 is advisory. The rules that bind a lawyer are those adopted in their licensing jurisdiction, so 512 is best read as a template that states have adapted — and in several cases (client consent, billing, vendor diligence) gone beyond.
Listed alphabetically. Under each state, instruments are grouped by type. The five duties — understand the tool, protect confidences, verify output, supervise, bill honestly — appear in nearly every instrument. The states diverge mainly on three points:
Client disclosure and consent. West Virginia requires informed consent confirmed in writing. Florida, California and North Carolina lean toward disclosure or consent for substantive AI use. Texas calls for case-by-case analysis without a blanket consent requirement. New York City says routine embedded tools (Word, Westlaw, Lexis) need no disclosure at all. Most others leave it to lawyer judgment.
Billing. States differ on how AI-driven time savings must be reflected in fees. Virginia LEO 1901 is the only opinion devoted entirely to the question.
Vendor diligence. States differ on how deeply lawyers must vet AI vendors' terms, data handling and security.
For firms practicing in multiple states, the conservative approach, as always, is to apply the strictest applicable state's rule — particularly on client disclosure — and document the choice-of-rule analysis.
Jump to a jurisdiction
The set of asks across states is mostly consistent: know what your tools do with client data, verify what they produce, write down how your people may use them, supervise that use, and bill honestly for the result. Each is a standing obligation, not a one-time project, and each is easier to meet with a managed intelligence provider — a partner like L2 Technology Solutions that selects, configures, governs and supports a firm's AI tools on an ongoing basis — than with ad hoc adoption.
Vendor diligence. Every opinion expects you to understand a tool's data retention, whether inputs train the model, its security posture and its terms of use before client information touches it. Reviewing that for each tool, and re-reviewing when terms change, is exactly what a managed provider does as a matter of course; L2 vets tools against the confidentiality standards in your state and deploys enterprise configurations rather than open consumer chatbots.
Written AI-use policy. California, Arkansas, Colorado and Rhode Island now treat a firm policy as an expected supervisory control. A managed intelligence provider like L2 can help you draft and update this policy over time.
Verification protocol. Florida, New York and Oklahoma have put citation verification into binding court rules, and sanctions for hallucinated authority are now routine. A managed provider builds verification into the workflow, so it does not depend on individual discipline, but also provides training for end users to ensure attorneys know their responsibility.
Supervision and training. Rules 5.1 and 5.3 make managing lawyers responsible for how everyone in the firm uses AI. L2 delivers the initial training and the periodic refreshers the bars now expect, tailored to the firm's tools and practice areas.
Client communication and engagement letters. Whether and how to disclose AI use depends on the strictest state touching a matter. A managed provider helps you decide what to say, draft the engagement-letter language and obtain specific consent where confidential data will be entered into third-party tools.
Agentic AI readiness. California and Oregon are the first movers on AI agents that draft, file or communicate with clients; more states will follow. We can help you stay on the forefront of the technology, as well as the policy when it comes to managed agents.
If you want help applying this guidance to your own practice — or a second opinion on a tool or policy you already have — reach out through the form below and we will follow up.
We typically respond within one business day.