Lawyers using AI: what the courts have sanctioned, what the bars allow, and what lawyers say they use

Most of what is written about lawyers using AI is either a warning or a sales pitch. This page is neither. It sets out what has actually happened, from the documents: the sanctions orders, with the court, the date and what the judge wrote; the bar opinions and court rules, quoted rather than summarised; and the surveys of what lawyers say they use. Then one problem none of those documents is about.

What courts have sanctioned and warned, dated

Four episodes account for most of the coverage. Each is described here from the court's own order or judgment, and only as far as that document goes.

New York, 22 June 2023: Mata v. Avianca

In Mata v. Avianca, Inc., 22-cv-1461 (S.D.N.Y.), Judge P. Kevin Castel's Opinion and Order on Sanctions of 22 June 2023 begins with the sentence every later authority quotes: "Technological advances are commonplace and there is nothing inherently improper about using a reliable artificial intelligence tool for assistance." It continues: "But existing rules impose a gatekeeping role on attorneys to ensure the accuracy of their filings." The order finds that Peter LoDuca, Steven A. Schwartz and the firm Levidow, Levidow & Oberman P.C. "abandoned their responsibilities when they submitted non-existent judicial opinions with fake quotes and citations created by the artificial intelligence tool ChatGPT, then continued to stand by the fake opinions after judicial orders called their existence into question."

The record in the order: Mr Schwartz researched and wrote the opposition brief using ChatGPT; Mr LoDuca signed and filed it without reviewing the authorities; when the court ordered copies of the cited decisions, the respondents produced fabricated "opinions" rather than withdraw them. The court found "bad faith on the part of the individual Respondents based upon acts of conscious avoidance and false and misleading statements to the Court." The sanction: "A penalty of $5,000 is jointly and severally imposed on Respondents," plus letters, within 14 days, to the client and to "each judge falsely identified as the author" of the six fake opinions. The court noted the firm had already arranged mandatory training on "technological competence and artificial intelligence programs" and declined to order more.

California, 12 September 2025: Noland v. Land of the Free

Noland v. Land of the Free, L.P., B331918, was decided by the Court of Appeal, Second Appellate District, Division Three, and certified for publication on 12 September 2025. The court called the underlying appeal "in most respects, unremarkable" and explained why it published anyway: "nearly all of the legal quotations in plaintiff's opening brief, and many of the quotations in plaintiff's reply brief, are fabricated." The count: "appellant's opening brief contains 23 case quotations, 21 of which are fabrications." Because "no California court has addressed this issue," the court wrote, "We therefore publish this opinion as a warning. Simply stated, no brief, pleading, motion, or any other paper filed in any court should contain any citations—whether provided by generative AI or any other source—that the attorney responsible for submitting the pleading has not personally read and verified."

The opinion records counsel's own account: attorney Amir Mostafavi "wrote initial drafts of the briefs, 'enhanced' the briefs with ChatGPT, and then ran the 'enhanced' briefs through other AI platforms to check for errors. Counsel admitted that he did not read the 'enhanced' briefs before he filed them." The court declined to allow corrected briefs, imposed a sanction of $10,000 "payable to the clerk of this court," directed counsel to serve the opinion on his client, and directed the clerk to serve it on the State Bar.

England and Wales, 6 June 2025: Ayinde and Al-Haroun

The judgment the IBA reported on 28 July 2025 as a UK judge's warning is Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin), handed down on 6 June 2025 by a Divisional Court of the High Court: the President of the King's Bench Division, Dame Victoria Sharp, sitting with Mr Justice Johnson. Two cases had been referred to it. In Ayinde, a judicial review against a London council, the claimant's barrister had cited five cases that did not exist; at an earlier hearing on 3 April 2025, Ritchie J had ordered her and the Haringey Law Centre "each to pay £2,000 to the defendant" as wasted costs and referred both to their regulators. In Al-Haroun, a damages claim, the solicitor's witness statement relied on research supplied by his own client; of "forty five citations that had been put before the court," the judgment says, "In eighteen instances, the case cited does not exist."

The court's general statement, at paragraphs 6 and 7, is the one lawyers in every jurisdiction have since been sent: such tools "are not capable of conducting reliable legal research. Such tools can produce apparently coherent and plausible responses to prompts, but those coherent and plausible responses may turn out to be entirely incorrect. The responses may make confident assertions that are simply untrue. They may cite sources that do not exist." And: "Those who use artificial intelligence to conduct legal research notwithstanding these risks have a professional duty therefore to check the accuracy of such research by reference to authoritative sources, before using it in the course of their professional work." On the solicitor, the court wrote that "A lawyer is not entitled to rely on their lay client for the accuracy of citations of authority." It found the threshold for contempt proceedings met in the barrister's case but decided not to initiate them, referred both lawyers to their regulators, and closed with a line aimed at the profession's bodies: "promulgating such guidance on its own is insufficient to address the misuse of artificial intelligence."

New York, 26 March 2025: the AI avatar in the Appellate Division

The episode CBS affiliates covered as an "AI-generated lawyer" was not a lawyer. As the Associated Press reported in a story carried by local stations on 4 April 2025, Jerome Dewald, who was representing himself in an employment appeal before the Appellate Division, First Department, had applied for permission to play a video for his argument on 26 March 2025 and used it to show an AI-generated figure reading his script. Justice Sallie Manzanet-Daniels stopped it: "It would have been nice to know that when you made your application. You did not tell me that sir." He was allowed to continue in person and later apologised to the court in writing. No lawyer was involved, so it says nothing about lawyers' professional duties; it is here because it is so often filed with the cases above.

How often this is happening

A Gizmodo article of 9 November 2025 ("Lawyers Are Using AI to Slop-ify Their Legal Briefs, and It's Getting Bad," widely reposted on Reddit) reports a New York Times story of 7 November 2025 on "a growing network of lawyers who track down A.I. abuses committed by their peers," and put the public database kept by Damien Charlotin at "509 cases so far." Read on 6 November 2026, the same database, which covers "all documents where the use of AI, whether established or merely alleged, is addressed in more than a passing reference by the court or tribunal," showed 2,266 cases, 1,305 attributed to self-represented litigants and 898 to lawyers, with entries dated as late as 3 November 2026. The decisions collected in Noland read the same way, among them federal and state courts in Alabama, Utah, Florida, California, New York, Missouri and Louisiana, from 2023 to 2025, each dealing with a filing that cited authorities nobody had opened.

What the ABA and state bars allow, as published

Bar guidance is not a permission slip, and this page does not read permission into it. What follows is what the documents say, quoted, with the date each was issued. All of it was re-read at the issuing body's own site on 6 November 2026.

ABA Formal Opinion 512, "Generative Artificial Intelligence Tools," 29 July 2024. The summary line: "To ensure clients are protected, lawyers using generative artificial intelligence tools must fully consider their applicable ethical obligations, including their duties to provide competent legal representation, to protect client information, to communicate with clients, to supervise their employees and agents, to advance only meritorious claims and contentions, to ensure candor toward the tribunal, and to charge reasonable fees." On competence: lawyers "need not become GAI experts" but "must have a reasonable understanding of the capabilities and limitations" of the tool, and such tools "cannot replace the judgment and experience necessary for lawyers to competently advise clients." On firms: "Managerial lawyers must establish clear policies regarding the law firm's permissible use of GAI, and supervisory lawyers must make reasonable efforts to ensure that the firm's lawyers and nonlawyers comply with their professional obligations when using GAI tools." Its confidentiality and informed-consent passages, which are what matter once client information is in play, are quoted and discussed in AI and attorney-client privilege.

The Florida Bar, Ethics Opinion 24-1, 19 January 2024. Headed "Advisory ethics opinions are not binding." Its opening sentence: "Lawyers may use generative artificial intelligence ('AI') in the practice of law but must protect the confidentiality of client information, provide accurate and competent services, avoid improper billing practices, and comply with applicable restrictions on lawyer advertising." It recommends "that a lawyer obtain the affected client's informed consent prior to utilizing a third-party generative AI program if the utilization would involve the disclosure of any confidential information," and treats the tool like a non-lawyer assistant: "a lawyer must verify the accuracy and sufficiency of all research performed by generative AI." Its conclusion: "a lawyer may ethically utilize generative AI technologies but only to the extent that the lawyer can reasonably guarantee compliance with the lawyer's ethical obligations."

The State Bar of California, Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law, 2026 revision. The current document says it "replaces the 2023 version and, at the request of the California Supreme Court, addresses the unique challenges presented by the use of agentic AI." Two lines carry most of its weight: "A lawyer's professional judgment cannot be delegated to AI," and, on tools connected to firm systems, "Lawyers must carefully evaluate and limit the scope of such access." It is quoted at length, alongside Opinion 512, in our practitioner's map of AI for lawyers, including the section on who in a firm decides.

The other states. Super Lawyers' "Can Lawyers Use AI in Court? State-by-State Rules" (by Eric Prindle, last updated 29 January 2026) lists guidance from eighteen jurisdictions: Arizona, California, the District of Columbia (Ethics Opinion 388), Florida (24-1), Kentucky (E-457), Massachusetts, Minnesota, Mississippi (Opinion 267), Missouri, New Jersey (the Supreme Court's preliminary guidelines), New Mexico, New York (the State Bar Association's task force report), North Carolina (2024 Formal Ethics Opinion 1), Oregon, Pennsylvania, Texas (Opinion 705), Vermont and West Virginia. Its own caveat applies to the list: "A majority of these publications are not formally binding but can generally be relied upon by lawyers practicing in these states." Each is a document to read at the bar's site, not a rule to infer from a table.

Court rules on AI in filings: Illinois and New York

Bars govern ethics; courts govern what may be filed. Two statewide court policies are short enough to quote.

Illinois Supreme Court Policy on Artificial Intelligence, announced 18 December 2024 and effective 1 January 2025. The court's release says the policy was adopted on the report of the Illinois Judicial Conference's Task Force on Artificial Intelligence. The policy's position: AI use is "authorized provided it complies with legal and ethical standards," and "Attorneys, judges, and self-represented litigants are accountable for their final work product." On disclosure, as the Illinois Supreme Court Commission on Professionalism quoted it on 10 January 2025: "Disclosure of AI use should not be required in a pleading."

New York, 22 NYCRR Part 161, "Use of Artificial Intelligence Technology," effective 1 June 2026 under Administrative Order AO/75/2026 of the Chief Administrative Judge (25 March 2026), for "all courts of the Unified Court System, in both civil and criminal cases." Section 161.3: "It is the policy of the Unified Court System that the use by attorneys and parties of artificial intelligence tools in preparing papers submitted to a court should not be prohibited, as long as such use is in accordance with the duties and responsibilities that apply to individuals who submit papers to a court. Since those duties and responsibilities already apply to all submissions, regardless of whether AI tools were used, attorneys and parties should not be required, upon submitting papers, to disclose to the court that they have used AI in the preparation of such papers." Section 161.4 lets any court adopt the model rule in Appendix A, under which a signature certifies that the paper "contains no such fabricated or fictitious content," and "If this court determines that this requirement has not been satisfied, such attorney or party may be subject to sanction or other remedial action." The definition of "paper" excludes evidence, which the rule says "are subject to separate considerations and requirements."

Neither rule disturbs an individual judge's standing orders. Before filing, the rule that binds you is the one for your court and your judge.

Can lawyers use ChatGPT? The rules, not a verdict

Read together, the documents above do not ban a named product, and they do not clear one either. They condition the use on three things the lawyer controls.

Whether your use is permitted is decided by your jurisdiction's rules, your court and your firm. This page reports the texts; it does not rule on them, and nothing on it is legal advice.

What lawyers say they use

The sanctions tell you what went wrong; the surveys tell you what is ordinary. The most-cited one is Clio's 2025 Legal Trends Report, the tenth edition. Clio's own release describes it as drawing on "aggregated and anonymized data from tens of thousands of legal professionals using Clio across the United States" plus survey and cognitive-study data; the report itself sits behind a form. As the Illinois Supreme Court Commission on Professionalism summarised it on 29 October 2025: "The report found that 79% of legal professionals use AI, the same percentage as in 2024," and "Clio found that only 40% of legal professionals are using legal-specific AI solutions, down from 58% in 2024." The uses it names are mundane: "AI used to summarize documents, prioritize tasks, or draft communications."

Put the two numbers together and the picture is this: most lawyers who use AI use a general assistant, not a legal platform, for drafting, summarising and correspondence. That is also what the court record shows. Gizmodo's gloss on the Times story, "Lawyers apparently love chatbots like ChatGPT because they can help them power through the drudgery of writing legal briefs," is a journalist's phrasing, not a survey finding, but the orders quoted above are consistent with it. For the platforms built for law, with their published prices, see AI for lawyers; for a firm's buying list, AI tools for law firms.

The client record: where Equerry fits

Everything above is about the law: research, citations, filings. There is a second body of work in a practice that none of these documents mention, and it is where the hours go between filings: what each client told you. What the general counsel said on the Tuesday call about the figure she would accept, which version of the term sheet their CFO sent on the 14th, what you promised before the board meeting, which matters they asked you to pause. That record is not legal research and no tool on this page is built to hold it. It lives in the partner's head, in the thread, and in notes across three apps, and it is why the first ten minutes of a client call go on reconstructing the last one.

Equerry (this is our app) is a personal assistant for iPhone and iPad that keeps that record and briefs you from it. It does not search the law, draft filings or check citations. You hand it things, the way you would hand them to an assistant, and ask about them later.

What you send Equerry, what you ask it, and what you get back
What you send The client side of a matter, kept apart from whatever you draft with. CC your Equerry address on the thread in which Marlow Freight's owner tells you which of the three claims he wants settled and which he will take to trial, or forward the email where his insurer's adjuster set the reserve figure. Share the mediator's scheduling letter as a PDF, or a photo of the timeline you drew with the client across a legal pad. Type a note after the client meeting recording that he will not give evidence about the second shipment. Record a voice note in the app on the walk back from the courthouse with what the judge said about the discovery deadline.
What you ask "Which of Marlow's three claims did he say he would settle, and what reserve did the adjuster put in writing?" "What did Marlow tell me he will not say in evidence?" "What did the judge set as the discovery deadline, and what did I tell Marlow I would file before it?"
What you get back The answer, drawn from what you sent it. When it draws on a note you saved, it carries a source chip: tap it to see the passage Equerry used and open the original. For a thread you CC'd or forwarded: What Equerry remembered from it is listed in Settings → Your memory, each line marked as from that email, and you can delete any line. If you allow calendar access and set up Meeting Prep, a short brief on where things stand with that client arrives before the meeting, built from what you have shared.

Equerry has an email address of its own, ending in @mail.getequerry.app. Put it on the CC line of a live thread and every later reply arrives on its own for as long as the address stays there; the original stays in your mailbox, and it never asks for your inbox. Nothing you keep in it reaches the tool you draft with, and nothing you draft with reaches it: the client record stays where you can see each line it kept and delete any line. Price: Pro is $39.99 a month or $329 a year.

Three cautions, in the spirit of the documents above. Everyone on a CC'd thread can see the address, so forward where that matters. What you send, with the calendar snapshot if you allow it, is processed by the outside providers listed on our AI disclosure page so that it can be searched and answered from; read that page the way Opinion 512 asks you to read any vendor's terms. And Equerry claims nothing about privilege or any bar rule. Whether client material goes into any tool on this page, ours included, is for your firm's policy, your jurisdiction's guidance and, where it applies, your client's informed consent. For what courts have said when AI and privilege meet, see AI and attorney-client privilege.

Frequently asked questions

Can lawyers use AI?

None of the documents quoted on this page bans use; each attaches conditions, and each says what they are in its own words. The ABA's Formal Opinion 512 (29 July 2024) opens by saying lawyers using generative AI "must fully consider their applicable ethical obligations"; the Florida Bar's Opinion 24-1 (19 January 2024) says lawyers "may use generative artificial intelligence" but "must protect the confidentiality of client information, provide accurate and competent services, avoid improper billing practices, and comply with applicable restrictions on lawyer advertising"; New York's Part 161 says use "should not be prohibited, as long as such use is in accordance with the duties and responsibilities that apply to individuals who submit papers to a court." What those duties require in your jurisdiction is a question for your state bar, your court and your firm's policy, not for this page.

Can lawyers use ChatGPT on client matters?

No bar opinion or court rule quoted on this page names a tool it bans. The conditions attach to what goes in and what comes out. On what goes in, ABA Opinion 512 says that before inputting information relating to a representation into a generative AI tool, lawyers "must evaluate the risks that the information will be disclosed to or accessed by others outside the firm." On what comes out, the Divisional Court in England and Wales said on 6 June 2025 that tools such as ChatGPT "are not capable of conducting reliable legal research" and that anyone using them has "a professional duty therefore to check the accuracy of such research by reference to authoritative sources." The two most widely reported sanctions, in New York in 2023 and California in 2025, both involved ChatGPT output filed without anyone checking the cases it cited. OpenAI's own terms for its personal and business plans are set out in our ChatGPT legal and ChatGPT confidentiality pages.

Which lawyers have been sanctioned for using AI?

The orders themselves name them. In Mata v. Avianca (S.D.N.Y., 22 June 2023), Judge P. Kevin Castel imposed a $5,000 penalty, jointly and severally, on Peter LoDuca, Steven A. Schwartz and the firm Levidow, Levidow & Oberman for a filing with six non-existent opinions generated by ChatGPT. In Noland v. Land of the Free (California Court of Appeal, Second District, Division Three, 12 September 2025), the court ordered attorney Amir Mostafavi to pay $10,000 to the clerk of the court after finding 21 of the 23 case quotations in his opening brief were fabricated. In England and Wales, a Divisional Court judgment of 6 June 2025 dealt with a barrister who had cited five cases that did not exist and a solicitor who filed a witness statement with 18 non-existent authorities among 45 citations; both were referred to their regulators. Damien Charlotin's public database listed 2,266 decisions involving AI-generated material when read on 6 November 2026, 898 of them attributed to lawyers.

Do state bars allow lawyers to use AI in court?

Bars publish ethics guidance; courts publish filing rules; the two are separate. On the bar side, Super Lawyers' state-by-state list (last updated 29 January 2026) collects guidance from eighteen jurisdictions and notes that "a majority of these publications are not formally binding." On the court side, two statewide rules are quoted on this page: the Illinois Supreme Court's policy, effective 1 January 2025, under which AI use is "authorized provided it complies with legal and ethical standards," and New York's Part 161, under which attorneys "should not be required, upon submitting papers, to disclose to the court that they have used AI," while a model rule lets any individual court require a certification that a paper "contains no such fabricated or fictitious content." Individual judges' standing orders can add requirements of their own, so the rule that applies is the one for your court.

What AI do lawyers actually use?

The most-cited survey is Clio's 2025 Legal Trends Report, which the Illinois Supreme Court Commission on Professionalism summarised on 29 October 2025: "79% of legal professionals use AI, the same percentage as in 2024," while "only 40% of legal professionals are using legal-specific AI solutions, down from 58% in 2024." In other words, most of the use is general assistants, not legal research platforms. The court record shows the same: the filings sanctioned in New York and California were drafted with ChatGPT. For a map of the tools by job, with prices, see our AI for lawyers guide.

Does Equerry do legal research or check citations?

No. Equerry is a personal assistant for iPhone and iPad that keeps the record of what each client told you, across the emails, documents and notes you choose to send it, and answers from that record with the source attached. It does not search case law, draft filings or verify citations, and it makes no claim about privilege, confidentiality rules or what any bar permits; whether client material goes into it is a decision for your firm's policy and your jurisdiction's guidance, the same as for any tool on this page.

Last updated 2026-11-06.

All product names, logos, and brands are property of their respective owners. Equerry is not affiliated with, endorsed by, or sponsored by any other product mentioned; comparisons are provided for informational purposes.