TL;DR: As of 2026, a growing list of US state bars has issued guidance on generative AI - California in 2023, Florida's Opinion 24-1, the New York State Bar's task-force report in 2024, and others including Texas, New Jersey, Pennsylvania, and the District of Columbia - under the umbrella of ABA Formal Opinion 512. The details differ, but the themes converge: competence, confidentiality, candor, supervision, reasonable fees, and client communication.
When a fast-moving technology meets a settled set of ethical rules, the bar's job is usually to explain how the existing rules apply rather than to write new ones from scratch. That is exactly what has happened with generative AI. Rather than banning it, US state bars have published opinions and guidance showing how a lawyer's established duties already govern its use. This article surveys where that guidance stands as of 2026. Treat it as a map and a starting point, not a substitute for reading the opinions that bind your own jurisdiction, because the landscape is still moving quickly and any snapshot dates fast.
Why are state bars weighing in on generative AI?
Two forces pushed the bars to act. The first was speed of adoption: within months of general-purpose chatbots becoming widely available, lawyers were using them for research, drafting, and summarizing, often without understanding how the tools worked. The second was a series of high-profile failures - lawyers who filed briefs containing confident, entirely fabricated case citations and were sanctioned for it. Those episodes, which we examine in our piece on AI hallucinations and Mata v. Avianca, made vivid the gap between what the technology appears to do and what it can be trusted to do.
It is worth being precise about what these documents are. Most state bar ethics opinions are advisory: they interpret the rules of professional conduct and tell lawyers how a disciplinary body is likely to view a practice, but they are not themselves binding rules. That distinction matters, yet it should not breed complacency, because opinions signal how the people who enforce the rules think. A lawyer who ignores a directly relevant ethics opinion is taking an avoidable risk. For the broader question of whether AI use is ethical at all, see our overview of AI ethics in legal practice under the ABA framework.
What does the ABA's Formal Opinion 512 establish?
The most influential single document is ABA Formal Opinion 512 on generative artificial intelligence tools, issued by the ABA Standing Committee on Ethics and Professional Responsibility in 2024. It does not break new doctrinal ground so much as walk methodically through the Model Rules and apply each to generative AI. It addresses competence under Rule 1.1, confidentiality under Rule 1.6, communication with clients under Rule 1.4, candor toward the tribunal under Rule 3.3, supervisory responsibilities under Rules 5.1 and 5.3, and fees under Rule 1.5.
The opinion's posture is permissive but cautious. It does not discourage lawyers from using generative AI; it stresses that they must understand the particular tool's benefits and limitations, take special care with confidentiality when a tool may learn from inputs, communicate appropriately with clients, and charge reasonable fees that reflect any efficiencies gained. Crucially, ABA opinions are advisory rather than binding - the rules that govern you are those adopted in your jurisdiction - so Opinion 512 is best read as a high-quality template that individual states adapt. With that umbrella in mind, the state-level picture comes into focus.
What have individual state bars said so far?
The table below summarizes the main US guidance as of 2026. It is necessarily compressed, and the emphasis column highlights what each body stressed most rather than capturing every nuance. Read the underlying documents before you rely on any of this for a live decision.
| Body | Guidance (as of 2026) | Key emphasis |
|---|---|---|
| State Bar of California | Practical Guidance for the Use of Generative AI (2023) | Confidentiality, competence, supervision, and fees |
| The Florida Bar | Ethics Opinion 24-1 on generative AI | Confidentiality, oversight of AI, fees, and advertising |
| New York State Bar Association | Task Force on AI report and guidelines (2024) | Competence, confidentiality, candor, and education |
| Texas, New Jersey, Pennsylvania, D.C. | State-specific opinions and guidance (2024 onward) | Competence, confidentiality, and supervision |
| American Bar Association | Formal Opinion 512 (2024) | National synthesis across the Model Rules |
California's practical guidance (2023)
The State Bar of California was among the first movers. Its Committee on Professional Responsibility and Conduct produced practical guidance on the use of generative AI in the practice of law, recommended in November 2023. The guidance is organized around existing duties rather than a new AI rule, and its most-quoted caution is about confidentiality: a lawyer should not input a client's confidential information into a generative AI tool that lacks adequate confidentiality and security protections. It also addresses competence, the duty to supervise, and the need to bill responsibly for AI-assisted work. The State Bar of California publishes its materials at calbar.ca.gov.
Florida's Ethics Opinion 24-1
The Florida Bar followed in early 2024 with Ethics Opinion 24-1, one of the first formal ethics opinions devoted to generative AI. Its framing is instructive: it treats generative AI somewhat like a nonlawyer assistant whose work the lawyer must supervise and remain responsible for. The opinion emphasizes protecting client confidentiality, exercising oversight over the tool's output, charging reasonable fees and costs, and taking care with AI in advertising and client intake - for example, chatbots that interact with prospective clients. The Florida Bar's resources are available at floridabar.org.
The New York State Bar's AI task-force report (2024)
New York took a broader, more report-driven approach. The New York State Bar Association convened a Task Force on Artificial Intelligence whose substantial report and guidelines were adopted in April 2024. Rather than a narrow ethics opinion, it offers a wide survey of AI's implications for the profession alongside practical guidelines, touching competence, confidentiality, candor, and - notably - lawyer education, recommending that practitioners build genuine understanding of the technology. The association's materials are published at nysba.org.
Texas, New Jersey, Pennsylvania, and the District of Columbia
Beyond the three above, several jurisdictions issued their own opinions or guidance from 2024 onward. Texas, New Jersey, Pennsylvania (including a joint effort with a metropolitan bar), and the District of Columbia have each addressed generative AI in some form. The specifics vary - some are formal ethics opinions, others are guidelines or preliminary statements - but they echo the same core duties of competence, confidentiality, and supervision. Because these documents are being issued and revised on a rolling basis, the responsible move as of 2026 is to check the current text directly with the relevant bar rather than rely on a secondhand summary.
What themes do the opinions share?
Read together, the opinions are more alike than different. Six themes recur across nearly all of them. Competence comes first: you must understand enough about a tool to use it responsibly, and you must verify its output rather than trust it - a discipline we detail in how to verify AI legal research. Confidentiality is a close second: do not feed client secrets into tools that lack adequate protection, and confirm a vendor does not train on your data. Candor toward the tribunal is the lesson of the citation scandals: never file what you have not checked. Supervision frames AI as something like nonlawyer assistance you must oversee, the subject of our guide to human-in-the-loop legal AI. Reasonable fees require that efficiencies from AI are not billed as if the work were done by hand. And client communication addresses when and whether to discuss AI use with clients. The emphasis shifts from state to state, but a lawyer who honors all six will be well inside the lines almost everywhere.
What do these opinions mean for your day-to-day practice?
Translating the guidance into habits is straightforward. Adopt a written approach to AI use - even a short internal policy - so decisions are consistent rather than ad hoc; our guide to AI governance for law firms shows how. Choose tools that do not train on your data and that can describe their security plainly. Verify every citation and factual claim against a primary source before it leaves your office. Supervise AI output the way you would a junior's draft, keeping your own judgment in the loop. Be transparent and fair about fees, passing genuine efficiencies to clients. And be mindful of the line between assisting your own practice and letting a tool give legal advice directly, which raises separate questions we cover in unauthorized practice of law and AI. Do these things and you are not merely compliant; you are using AI the way the bars clearly hope lawyers will.
How does Judicio map to these expectations?
It is fair to ask how a specific platform measures against this guidance, and the honest answer is that a tool can support your duties but never discharge them. With that caveat, Judicio is built around exactly the obligations the opinions stress. Competence and candor are served by citation discipline: every answer, finding, and date in Legal Research carries the exact page and the quoted passage, the labels are deterministic rather than AI-generated, web sources are archived as permanent PDFs, and you can export an evidence pack - so verification, the step the bars insist on, takes seconds. Confidentiality and supervision are served by the data model: Judicio does not train on your uploads, hosts on Google Cloud Platform, and provides role-based access with a full audit trail, while one upload into the File Library keeps you in control of where a document goes. None of this removes your responsibility - outputs are explicitly not legal advice, and you remain the lawyer of record. For the underlying ethics framework, see is AI ethical in legal practice.
How do you stay current as guidance evolves?
The single most important thing to remember about this survey is that it is dated the moment it is written. New opinions appear, existing ones are revised, and the ABA and state bars continue to refine their positions. Subscribe to updates from your own jurisdiction's bar, re-check the guidance before you rely on it for an important decision, and treat any 2026 snapshot - including this one - as a prompt to confirm rather than a final word. You can put the practices the opinions recommend to the test in Judicio with a 7-day free trial: 500 credits, no credit card required, with Professional plans at $200 per month for 5,000 credits. For a walkthrough tailored to your firm's policies, contact us.
This article summarizes publicly available guidance as of 2026 and is not legal or ethics advice; AI outputs are not legal advice, and you remain responsible for compliance in your own jurisdiction.