AI in Legal Practice: State-by-State Regulation and Bar Guidance (2026)

AI legal regulation by state is a patchwork, not a single rule. ABA Formal Opinion 512 (July 29, 2024) is the closest thing to a national baseline. On top of it, a growing number of state bars have issued ethics opinions or guidance on lawyers using generative AI, and a growing number of state courts have adopted AI policies or filing rules of their own. Where a state has said nothing specific, the rules of professional conduct it already had still apply.

Reviewed and corrected on 12 October 2026. This is general information, not legal advice; check the cited opinion, rule or order at its source before relying on it.

If you run a legal team, the practical question is narrow. What has your state bar said, what has your state's highest court adopted, and what have the judges you appear before ordered? Those are three different layers, and the court layers moved faster in 2026 than the bar layer.

A word on method. AI-written trackers are known to repeat opinion numbers that do not exist. Every row below comes from a document we read on the issuer's own site (a bar, a court, a legislature) in October 2026. Where we could only locate a document, or could not read it, the row says so. A state that is missing from the tables is not a state with no AI guidance. It means we did not confirm one.

TL;DR

• ABA Formal Opinion 512 (July 29, 2024) is the baseline. It applies the existing Rules of Professional Conduct to generative AI: competence, confidentiality, communication, candor, fees and supervision. It is persuasive guidance, not law. Most state opinions follow its structure.
• Bar guidance comes in different strengths. Formal ethics opinions (for example Florida 24-1, Texas 705, Oregon 2025-205, Washington 2025-05, Alaska 2025-1, New Mexico 2024-004) sit next to a Virginia opinion that only covers fees, a Missouri informal opinion, an Ohio staff guide, a Georgia toolkit that says it is not an opinion, and a Rhode Island court order that makes its guidelines advisory.
• The same duties repeat. Understand the tool, protect client confidences, verify every output, bill honestly, and supervise the AI like a junior associate.
• Courts started writing binding filing rules in 2026. Florida's Supreme Court amended Rule 2.515 so that every signer of a filing represents that the legal authorities in it exist and are accurately cited (effective June 15, 2026). Connecticut's Practice Book section 4-9 took effect June 23, 2026 and requires anyone who uses generative AI in a filed document to verify the citations, authorities and evidence. The Oklahoma Court of Criminal Appeals adopted Rule 1.17 on February 18, 2026. In federal court, the Northern District of Texas requires a first-page disclosure under Local Rule 7.2(f).
• Two states went past opinions in 2026. The Colorado Supreme Court amended its Rules of Professional Conduct on January 8, 2026, adding Scope paragraph [20A] and Comment [9] to Rule 1.1 on technology and AI. California enacted SB 574 on September 30, 2026. It adds Business and Professions Code section 6068.1, effective January 1, 2027, which requires an attorney who uses generative AI to take reasonable steps to verify outputs and to disclose that use to the court for documents submitted to it.
• Many court policies cover only the court's own staff and judges. California's Rule of Court 10.430, the New York and Pennsylvania court-system policies, Delaware's interim policy, South Carolina's interim policy, Arkansas Administrative Order 25, North Dakota Administrative Order 36 and Massachusetts's interim guidelines are examples. Check whose conduct a policy actually regulates.
• The sanctions risk is the verification rule. Fake AI-generated citations have drawn sanctions since Mata v. Avianca in 2023. Our sanctions tracker keeps the running list.

The baseline every lawyer is under

Start with the document that sits under all the others. ABA Formal Opinion 512, issued July 29, 2024 by the ABA Standing Committee on Ethics and Professional Responsibility, is the ABA's guidance on lawyers using generative AI. It does not create new rules. It reads the rules you already have onto a new tool.

Six duties carry the opinion, and they show up again in most state opinions below:

1. Competence (Rule 1.1). Know what the tool can and cannot do before you use it on client work, and keep up as it changes.
2. Confidentiality (Rule 1.6). Check how a tool stores and trains on your inputs before you put client information into it.
3. Communication (Rule 1.4). Tell the client when your use of AI is material to the work or the fee.
4. Candor and meritorious claims (Rules 3.1, 3.3 and 8.4(c)). Verify every citation and quote before it goes in a filing.
5. Reasonable fees (Rule 1.5). Bill the time you actually spend, and explain any charge for a per-use tool before you bill it.
6. Supervision (Rules 5.1 and 5.3). You are responsible for what the AI produces the way you are responsible for a paralegal's work.

The ABA opinion is persuasive, not binding. For how that works, see are ABA formal opinions binding, and for a full walkthrough of the duties, the ABA Formal Opinion 512 guide. For the broader compliance picture (state AI statutes, the EU AI Act, the federal framework), see our AI legal regulation state and federal update.

The authority hierarchy (read this before the table)

These documents do not carry equal weight. Before you trust any single entry in the tracker, place it on this ladder, from least to most binding:

1. ABA formal opinion. Persuasive only. It binds no lawyer until a state adopts it or a court relies on it.
2. State-bar ethics opinion or guidance. Advisory in most states. Florida's Opinion 24-1 opens with "Advisory ethics opinions are not binding," Kentucky's says formal opinions are advisory only under Supreme Court Rule 3.530, and West Virginia's Office of Disciplinary Counsel says its Legal Ethics Opinions are advisory only and not binding on the Supreme Court of Appeals. Virginia is a special case: its fee opinion, LEO 1901, was approved by the Supreme Court of Virginia on November 24, 2025, effective immediately. We did not find a statement of the legal effect of that approval, so do not assume more than "approved by the court."
3. State supreme court or court-system rule. Binding on the people it covers, and that group differs. Connecticut's Practice Book section 4-9 and the Oklahoma Court of Criminal Appeals's Rule 1.17 reach lawyers who file documents. California's Rule 10.430 and Delaware's interim policy reach court staff and judicial officers. Colorado's January 8, 2026 amendment is different again: the state supreme court changed its Rules of Professional Conduct themselves (Scope paragraph [20A] and Rule 1.1 comments [8] and [9]), so it reaches every Colorado lawyer, but it adds no AI-specific filing or disclosure step. Where a legislature has acted, a statute adds another layer that binds lawyers directly: California's SB 574 (section 6068.1), effective January 1, 2027.
4. Individual judge standing order or local court rule. Binding in that courtroom or district. It can require a certificate or disclosure your bar does not, and it controls your filing.

Here is the opinionated part, because the ladder alone does not tell you how to act. Treat two things as operationally mandatory even where they are not technically binding: your own state bar's AI guidance, because it is what a disciplinary panel will measure you against, and any order or local rule from the court you are appearing in. Ignore neither.

What should not drive your firm-wide policy: another state's toolkit or informal column, and the ABA opinion read alone. They are useful reference. Build to your own jurisdictions plus the strictest verification rule you find.

Methodology

For each jurisdiction we looked for four kinds of document: a bar ethics opinion or guidance on generative AI, a state court-system AI policy or rule, a bar or court task force report, and court rules or orders on AI in filings. A document counted only if it names generative or artificial intelligence and says how lawyers (or judges and court staff) may use it. We read each one on the issuer's site. Where a bar's or court's own page was blocked to us, we used a text copy of that same page and say so in the notes. Third-party trackers were used only to find leads.

We did not count generic technology-competence rules, CLE programs, or articles that only summarize the ABA opinion. The tables tag every entry by type so you do not over-read a toolkit or a column.

Dates and status are as of October 12, 2026. This is a snapshot of fast-moving material: several of the court items below took effect in 2026.

AI legal regulation by state: the 2026 tracker

Bar opinions, guidance and task force reports we read

Court rules, policies and orders we read

What we could not confirm

We located the following but did not read the primary document, or could not read it in full, so they are not tabled. Not verified in our October 2026 review. Check each issuer's own site.

• Mississippi. A Mississippi Office of State Public Defender training presentation cites Mississippi Bar Ethics Opinion No. 267 together with ABA Formal Opinion 512 on generative AI. The Bar's page for the opinion shows no text (the Bar says its formal opinions are available through vLex Fastcase), so we did not confirm its date or holding.
• Kansas, Maryland, South Dakota, Utah. Kansas Supreme Court Administrative Order 2025-CM-017 (dated Feb 27, 2025, read on kscourts.gov) creates an ad hoc AI committee to recommend policies for court employees and for attorneys, parties and the public, and we did not confirm a resulting policy. The Utah courts' HR page says Judicial Council interim rules on generative AI apply from Oct 27, 2025 to court-related work and court-owned devices and prohibit any AI tool they do not expressly permit; we did not read the rules themselves. The Maryland Judiciary's information security policy (January 2026) refers to Judiciary AI guidelines that we did not read. For South Dakota we have only tracker reports of Unified Judicial System guidance, which we did not find.
• Tennessee. The Tennessee Bar Association lists an AI Task Force on tba.org. We found no bar ethics opinion or adopted court rule on AI for lawyers.
• Louisiana. The Supreme Court's 2024 letter to the bar is in the table above. We did not check Louisiana's 2025 evidence-authenticity law.

States where our review did not confirm anything

For Idaho, Maine, Montana, Nebraska, Nevada, Wisconsin and Wyoming, our searches did not turn up a bar ethics opinion or a court AI rule that we could read, but we did not check every page of each bar's and court's site, so this is not proof that nothing exists. Not verified in our October 2026 review. Check the state bar's ethics-opinion page and the judiciary's rules and orders page. A few leads we saw but did not tabulate: the Nevada courts' Administrative Office publishes an "Artificial Intelligence Guide" for judicial officers, and the Wisconsin court system's rule-petition pages show that petition 26-02 (filed February 19, 2026, proposing a new SCR chapter 76 on generative AI in court proceedings) was followed by a September 30, 2026 letter from the petitioner asking to withdraw it, and that a different petition, 26-06 (filed September 28, 2026, proposing SCR 76.20 to 76.24 on equal access to AI tools in legal proceedings), is pending. Neither is a rule. See petition 26-02 and petition 26-06. Many bars have run CLE programs and published articles or president's columns on AI. Those are not opinions. The ABA 512 duties and each state's rules of professional conduct apply in all of them.

What changed in 2025 and 2026

The first wave (2023 into early 2024) was advisory bar opinions: Florida, Kentucky, Pennsylvania, DC, the NYC Bar and the New Hampshire Ethics Committee, plus New Jersey's court guidelines and a Louisiana Supreme Court letter. They said the existing rules apply. The newer wave differs in three ways.

More bars issued opinions. Texas (Opinion 705, February 2025), Oregon (2025-205, February 2025), Alaska (2025-1, April 2025), Washington (2025-05), North Carolina (adopted November 1, 2024), New Mexico (September 2024), West Virginia (2024) and Alabama (2026) added bar guidance, and Virginia's Supreme Court approved a fee opinion in November 2025. In 2026 the Ohio staff guide and Rhode Island's Supreme Court order followed, Colorado's Supreme Court amended its Rules of Professional Conduct in January, and California replaced its 2023 guidance in May and enacted SB 574 in September.

Court rules reached filers. Florida's amended Rule 2.515 (effective June 15, 2026), Connecticut's section 4-9 (June 23, 2026) and Oklahoma's Rule 1.17 (February 18, 2026) put a verification duty into the court's own rules, with sanctions attached. North Dakota's notice says disclosure is not required solely because AI was used, and Illinois's policy says disclosure should not be required in a pleading. Courts are not uniform on disclosure. California went further by statute: SB 574, effective January 1, 2027, requires disclosure of generative AI use to the court on documents submitted to it (Business and Professions Code section 6068.1(a)(3)(C)) and bars filed papers from containing citations the responsible attorney has not personally verified (Code of Civil Procedure section 128.7(b)(2), as amended).

Many court policies are for the court's own people. California's Rule 10.430, Delaware, New York, Pennsylvania, South Carolina, Massachusetts, Arizona, Arkansas, Iowa and North Dakota's Order 36 govern judges and court staff. They tell you how the court uses AI, not what you must file.

So the clean split today: bar opinions are mostly advisory and tell you how to practice, while court rules, local rules and individual judge orders are binding on the filings they cover, and California's statute will add to that from 2027. The advisory layer is broad. The binding layer is newer, narrower, and moving.

The common throughline

Read a dozen of these opinions back to back and the wording barely changes. You do not need 50 different policies, you need one that hits the points the opinions make. Five duties recur.

Competence. The opinions tie AI to Rule 1.1. You have to understand the tool well enough to use it responsibly, including its biggest failure mode: it produces confident, fluent, wrong answers. DC's opinion and Oregon's both point to a Stanford finding that general-purpose models hallucinated at least 75% of the time on questions about a court's core ruling, which is why competence and verification travel together.

Confidentiality. The recurring warning is about inputs. Putting client information into a tool that retains or reuses it can breach Rule 1.6. Florida recommends client consent before using a third-party tool that receives confidential information. Oregon requires informed consent to use client information in an open model. North Carolina says lawyers should avoid putting client-specific information into publicly available AI resources. Check the terms and the data handling first.

Candor and verification. This is the one with teeth. Texas Opinion 705, Pennsylvania 2024-200, North Carolina's 2024 FEO 1 (which ties a signed pleading to Rule 3.1 and Rule 11), Ohio's guide and Connecticut's rule all put the duty on the lawyer: the tool does not carry it. For how to do it, see how to verify AI legal citations before filing.

Reasonable fees. Two halves. You cannot bill a client for hours you did not work. Texas, North Carolina and Alaska say so in terms, and Oregon says that where AI saves significant time, lawyers may not duplicate charges or falsely inflate billable hours. And Virginia's LEO 1901 adds the flip side: Rule 1.5 does not equate reduced time with proportionally reduced fees, so a reasonable non-hourly fee need not shrink just because AI made you faster. Kentucky, by contrast, says a lawyer must consider reducing fees when AI cuts the time.

Supervision. Treat AI like a junior associate whose work you sign. You own the output under Rules 5.1 and 5.3. Alabama, Florida and Texas all compare the tool to a nonlawyer assistant.

Court-level rules and standing orders

The bar opinions tell you how to practice. The courts tell you what to file, and that layer moves faster and varies more.

The standing-order trend started with one judge, and has since moved into local rules. On May 30, 2023, Judge Brantley Starr (N.D. Tex.) posted the first standing order requiring attorneys to certify either that no portion of a filing was drafted by generative AI or that a human checked any AI-drafted language for accuracy (archived copy of his page). His page on the court's site now says he no longer requires that certification in light of the court's adoption of Local Civil Rule 7.2(f) and Local Criminal Rule 47.2(e). The district judges proposed both rules in Special Order 2-97 (February 20, 2024, with public comment open until June 3, 2024) and adopted them in Special Order 2-98 (June 2024), effective September 3, 2024. Rule 7.2(f), which appears in the Local Civil Rules effective September 1, 2026, requires a brief prepared using generative AI to say so on its first page under the heading "Use of Generative Artificial Intelligence," and the presiding judge may direct the filer to identify the specific parts. A party whose brief has no such disclosure certifies that no part of it was prepared using generative AI. Other judges and courts have their own orders. Ropes & Gray said in May 2026 that its AI court order tracker catalogs over 550 standing orders, local rules and decisions on generative AI in state and federal courts, and the tracker has a filter for orders that require disclosure and/or verification. They are not uniform: some demand a certificate, some disclosure, some only verification, and many courts have nothing.

The Fifth Circuit declined a special rule. It proposed a certificate requirement in late 2023, took comment, and in June 2024 posted a notice that it had "decided not to adopt a special rule regarding the use of artificial intelligence in drafting briefs at this time." Its reasoning is the throughline of this whole topic: existing rules already require accurate filings, and "'I used AI' will not be an excuse for an otherwise sanctionable offense."

State courts are now writing rules. Florida's Rule 2.515(d)(2), Connecticut's section 4-9 and Oklahoma's Rule 1.17 are verification rules for filers, and Florida's administrative order AOSC26-12 stops Florida courts from adding their own AI disclosure or certification requirements. North Dakota's June 2026 notice and Illinois's policy are on the permissive side and say disclosure is not required merely because AI was used. Several other courts (California, Delaware, New York, Pennsylvania, South Carolina, Massachusetts, Arizona, Arkansas) regulate only their own judges and staff.

The enforcement risk is the verification rule, not an AI rule. Sanctions for fake AI citations trace back to Mata v. Avianca (S.D.N.Y., June 22, 2023), where Judge Castel imposed a $5,000 penalty, jointly and severally, on the two lawyers and their firm for filing ChatGPT-invented cases. The cases have not stopped. For the running list, see our AI hallucination sanctions tracker and the deeper AI hallucinations and sanctions piece.

How a multistate team should operate

If you sit in-house or run a small firm across jurisdictions, the wrong move is 50 separate playbooks. The opinions overlap enough that three artifacts cover the field.

One AI-use policy, written to the duties. Anchor it to the six ABA 512 duties and the strictest verification language in any state you touch. A policy that requires human verification of every AI output, keeps client confidences out of open tools unless the client has given informed consent, and assigns a named owner is consistent with the opinions we read. Update it on a fixed cadence, not ad hoc.

One vendor-review checklist, run before any tool is approved. The opinions that go past competence (North Carolina, Washington, Alabama, Oregon) push diligence on the tool. Your checklist: does it train on or retain your inputs, where is the data stored, is there a confidentiality and security commitment in the contract, and what can you check about its outputs. Read each vendor's own current terms for your tier, because the data terms differ by product and plan. Approve tools centrally so individual lawyers are not each negotiating terms.

One court-rule check, run per filing. This is the step teams skip. Before every filing, check the local rules and the standing orders for that specific court and judge, because the disclosure or verification requirement lives there, not in your state bar opinion. Build it into the filing checklist the way you already check page limits and formatting.

Worked example: a state-court pleading in Illinois, a federal brief in Texas

Say your team uses AI help on two filings. The first is a pleading in an Illinois state court. The Illinois Supreme Court's policy says disclosure of AI use should not be required in a pleading, so no AI statement is needed there. You still have to review the AI content and you are accountable for the final work. The second is a brief in the Northern District of Texas. Illinois's policy does not govern a federal court in Texas. Local Rule 7.2(f) does, and it requires the first-page statement.

The right sequence: confirm the rules of the court before drafting, not after. Draft to that court's format, run your own citation check against a real database (not the AI's own claim that the cite is real), and attach whatever the court requires. The state policy told you what Illinois asks. The Texas local rule told you what to put on page one. Both matter, and the one for the court you are in is the one that gets your brief stricken or sanctioned if you miss it.

Put the three artifacts in place and the patchwork stops being 50 problems. It becomes one policy, one checklist, and one per-filing habit that holds up as more states act.

FAQ

Is it legal for lawyers to use AI in 2026? Yes. None of the bar opinions, court rules or policies we read bans lawyers from using generative AI. They permit it subject to the existing duties of competence, confidentiality, candor, fees and supervision. The constant condition is that you verify the output.

Which states have issued AI ethics opinions for lawyers? We read bar opinions or guidance from Alabama, Alaska, California, DC, Florida, Georgia (a toolkit), Kentucky, Minnesota (an informal column), Missouri (an informal opinion), New Mexico, New York (NYC Bar opinions and a NYSBA task force report), North Carolina, Ohio (staff guidance), Oregon, Pennsylvania, Rhode Island (court-adopted guidelines), Texas, Virginia (fees only), Washington and West Virginia. Colorado went further and amended its Rules of Professional Conduct (January 2026). Louisiana's Supreme Court sent the bar a letter (January 2024), and New Hampshire's Ethics Committee published three Ethics Corner articles in 2024. Mississippi has an Ethics Opinion No. 267 that we could not read. Counts differ by what a tracker counts, so check your own bar.

Is ABA Formal Opinion 512 binding? No. ABA formal opinions are persuasive, not binding. They matter when a state adopts them or a court or disciplinary body relies on them. Most state opinions track its framework, so following it is the safe default, but your own state's rules control.

Do I have to tell the court I used AI? It depends on the court. Some courts require a statement: the Northern District of Texas requires a first-page disclosure for briefs under Local Rule 7.2(f). Some require verification instead: Florida (every signer represents that the cited authorities exist and are accurately cited), Connecticut and the Oklahoma Court of Criminal Appeals. Some say disclosure is not required merely because AI was used: Illinois's policy and North Dakota's notice. New Jersey's guidelines say the Rules of Professional Conduct do not require disclosing AI use. California's SB 574 will require disclosure of generative AI use to the court on documents submitted to it, effective January 1, 2027. Always check the rules of your court and the orders of your judge before filing.

What is the most common requirement across all of these? Verification. The opinions and rules we read put the duty to check AI-generated citations, quotes and analysis on the lawyer. It flows from the duty of candor to the tribunal and is the rule that draws sanctions when ignored.

Can I bill a client for time saved by AI? Not as hours you did not work. Texas and North Carolina say you may not bill hourly clients for time the AI saved, Alaska says you may not bill for time you did not work, and Oregon says you may not duplicate charges or falsely inflate hours. Virginia's LEO 1901 says a reasonable non-hourly fee need not shrink just because AI made you faster, and Kentucky says a lawyer must consider reducing fees when AI reduces the time. Whether and how to pass through a tool's cost is also covered in several opinions, so read your own state's.

What happens if AI invents a case in my brief? You can be sanctioned. In Mata v. Avianca (S.D.N.Y. 2023) the court imposed a $5,000 penalty on two lawyers and their firm, and courts and disciplinary bodies have continued to act since. The duty to verify is the lawyer's, not the tool's.

My state is not in the tables. Does that mean AI is unregulated for lawyers here? No. It means we did not confirm an AI-specific document there in our October 2026 review. Your state's rules of professional conduct and the court rules where you file still apply, and the ABA 512 framework is a sensible baseline.

For the US statutes, regulations and court rules these laws sit in, see the US primary law API.

Arshita Anand

Co-Founder & CEO · Attorney

Arshita leads product and strategy at Vaquill AI, the US primary law data API. As an attorney, she decides which law we carry, how each section is sourced and cited, and what legal AI and compliance teams need from the data.