AI in law: how courts and lawyers use it, and how the law regulates AI

The short version

• AI in law supports work such as research, document review, summaries and first drafts, and the lawyer or judge stays answerable for the result.
• Lawyers may use AI under their existing professional duties, and from 1 January 2027 a California statute adds AI duties for attorneys.
• Court guidance we checked lets judges use AI for support work but keeps the decision with the judge.
• Courts treat invented citations as the lawyer’s failure: in Mata v. Avianca the lawyers and their firm were fined US$5,000.
• From 2 December 2027, the EU AI Act can treat AI that helps a court or an alternative dispute resolution (ADR) body decide a case with legal effects as high-risk.
• UK and US courts require a human inventor, US copyright needs a human author, and a Canadian tribunal held an airline liable for what its chatbot said.^{1,2,4,5,6,7,8,9,12,13,15,16,17,18,19}

How is AI used in law today?

Artificial intelligence (AI) in law is used to support tasks such as research, document review, due diligence, summarising, transcription and first drafts, the Law Society of England and Wales says, and lawyers stay answerable for the result.¹

In the American Bar Association’s 2024 Legal Technology Survey, 30.2% of lawyers said their offices used AI-based tools, rising to 47.8% at firms of 500 or more lawyers. The benefit most often picked was “saving time/increasing efficiency” (54.4%). Our legal tech guide covers the tools lawyers use, adoption figures and buying advice.³

For judges, guidance in England and Wales calls summarising, presentations and administrative tasks “potentially useful” uses of generative AI, which writes new text, and marks legal research and analysis “not recommended” (p. 7).¹⁹

That guidance and the American Bar Association’s Formal Opinion 512 name the same main risks. Tools can invent cases, citations and quotations, known as hallucination. Outputs can carry bias from training data, and tools that learn from what users enter can disclose client information (judicial guidance, s 3; ABA Formal Opinion 512, pp. 3, 6 and 7).^2,19

Is it legal and ethical for lawyers to use AI?

Yes. Lawyers may use AI in the jurisdictions we checked (how we research), as long as they keep to their existing professional duties. ABA Formal Opinion 512 (29 July 2024), the American Bar Association’s ethics guidance, applies the duties of competence, confidentiality, client communication, candour, supervision and reasonable fees to generative AI.^1,2

California’s SB 574, approved on 30 September 2026, takes effect on 1 January 2027. An attorney “shall not delegate the practice of law to generative artificial intelligence”, must take reasonable steps to verify outputs, including all case and statutory citations, and must disclose AI use to the court (Business and Professions Code s 6068.1(a)(2), (a)(3)(B), (C)).^4,5

Our legal tech guide sets out lawyers’ professional duties when using AI, side by side, including the EU AI Act’s duty on AI literacy. Our commentary covers how India’s courts are handling AI in legal work.

Can judges and courts use AI?

Yes. Judges may use AI for support work, under court guidance that keeps the judge responsible for the decision. The judiciary of England and Wales says “AI tools must not determine the outcome of a case” (Position Statement of 9 October 2026, para 10). Its judicial guidance of 31 October 2025 makes judges “personally responsible for material which is produced in their name” (p. 6).^6,19

The same guidance warns of deepfakes (AI-faked images, audio or video) and “white text”, hidden prompts that a person cannot see (p. 7). The Illinois Supreme Court says judges “remain ultimately responsible for their decisions” (policy, p. 2).^7,19

In the EU, the AI Act can treat AI used by or for a judicial authority to research and interpret facts and the law and to apply the law to facts as high-risk (Annex III point 8(a)). UNESCO published Guidelines for the use of AI systems in courts and tribunals in 2025. The Council of Europe, a human rights body separate from the EU, adopted its European Ethical Charter on the use of AI in judicial systems on 3 and 4 December 2018.^12,21,22

One judge’s AI errors reached a US appeal court. In Jackson Federation of Teachers v. Fitch, a federal trial judge’s temporary restraining order (a short-term court order) referred to parties and declarations not in the record, and the corrected order, later folded into a preliminary injunction, cited a case that does not exist (pp. 3, 6 to 8). The judge had said a law clerk used AI on an earlier draft (p. 7). In an opinion of 22 October 2026, the US Court of Appeals for the Fifth Circuit vacated the injunction, also for legal error, and directed that another judge take the case (pp. 2, 4, 27). That is the position as of 23 October 2026; we could not check the court’s docket for a petition for rehearing.²³

Proposed US Federal Rule of Evidence 707, still under consideration in May 2026, would apply the Rule 702 reliability test for expert evidence to machine-generated evidence that would face that test if a witness gave it (September 2026 report, Rules p. 26).^24,25

The court rules we checked take three approaches to AI in filings: disclosure of AI use, certification that every citation exists, or reliance on existing duties.^7,11,26

What happens when lawyers cite fake cases that AI invented?

Courts treat a fake citation that AI invented, known as a hallucination, as the lawyer’s failure. We found no official count, but the High Court of England and Wales said in 2025 that there had been “many instances, in countries around the world” ([83]).³²

In Ayinde, [2025] EWHC 1383 (Admin), the High Court of England and Wales held that lawyers must check AI research against authoritative sources ([7], [8]). It sent the judgment to the Bar Council, the Law Society and the Council of the Inns of Court ([82]). Our explainer covers how English courts have handled false citations.³²

In Mata v. Avianca (22 June 2023), a New York federal court found nothing “inherently improper” in using a reliable AI tool but fined the lawyers and their firm US$5,000 for submitting non-existent opinions. A California appeal court published Noland v. Land of the Free, L.P. (12 September 2025) “as a warning” (p. 2), fined the lawyer US$10,000 and sent the opinion to the State Bar (pp. 30, 32).^9,10

Court rules are changing. Since 15 June 2026, Florida’s Rule 2.515(d)(2) makes whoever signs a filing represent that the legal authorities identified “exist and are accurately cited” (SC2026-0673, 28 May 2026); California and New South Wales have similar citation duties (see the table above). In the UK’s Upper Tribunal, an immigration judicial review claim form now needs a statement of truth that each authority cited exists, can be found from its citation and supports the point it is cited for ([2026] UKUT 81 (IAC), headnote 3).^4,11,29,33

The US federal Civil Rules Advisory Committee has not followed: it removed proposals on AI “hallucinations” from its agenda (September 2026 report, Rules pp. 17, 18). Arbitrators are not exempt either: in April 2026 the Superior Court of Québec annulled an arbitral award that relied on non-existent authorities (ARIHQ c. Santé Québec, 2026 QCCS 1360; the Québec annulment explained).^25,34

Will AI replace lawyers or judges?

No rule we checked lets AI replace a lawyer or a judge: each keeps a named person answerable for legal advice, filings and decisions. California applies this rule of non-delegation to attorneys, as above, and from 1 January 2027 to arbitrators, who may not delegate “any part of their decisionmaking process” to generative AI (Code of Civil Procedure s 1282.1(b)(1)).^4,5

In May 2025 the Solicitors Regulation Authority (SRA) authorised the first purely AI-based law firm it has approved, for debt recovery through the small claims process. The SRA said the system “is not autonomous” and that named regulated solicitors stay accountable (6 May 2025).³⁵

The US Chief Justice wrote in 2023 that “human judges will be around for a while”, but also that judicial work “will be significantly affected by AI” (2023 Year-End Report, p. 6). Whether AI can be an arbitrator is covered in our arbitration guide.³⁶

How is AI used in dispute resolution?

In arbitration and other dispute resolution, the AI uses we found support the people who decide. The American Arbitration Association (AAA) offers an opt-in AI arbitrator that drafts awards in two-party, documents-only construction cases, and a human arbitrator reviews, may rewrite and signs each award (from 3 November 2025; R-8). From 2 December 2027, the EU AI Act can treat AI that helps a court or an ADR body decide a case with legal effects as high-risk (Annex III point 8(a); Regulation (EU) 2026/1744, art. 1 point (40)(b)).^12,13,37,38

Separate guides cover AI in arbitration and AAA’s AI arbitrator, AI in mediation, online dispute resolution and ombudsman schemes, and AI used to prevent disputes before they arise.

How is AI itself regulated?

AI itself is regulated unevenly: the EU has one comprehensive law, the Council of Europe a treaty, and the United States mainly executive orders and state laws. The EU AI Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024 and applies generally from 2 August 2026 (art. 113; Regulation (EU) 2026/1744, recital 1).^{12,13,14,39,40}

Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on 27 July 2026. It moved the high-risk obligations to 2 December 2027 for the uses listed in Annex III, which include AI that helps a court or an ADR body decide a case with legal effects, and to 2 August 2028 for AI in products under the EU laws listed in Annex I (art. 1 point (40)(b)). High-risk systems for public authorities already on the market or in service by then have until 2 August 2030 (art. 1 point (39)(a)).¹³

The Council of Europe’s Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (CETS No. 225, 5 September 2024) applies to AI used by public authorities and those acting for them; each party must also address private actors’ AI and declare how (art. 3(1)). It enters into force about three months after five signatories, including at least three Council of Europe member states, agree to be bound (art. 30(3)). The EU approved it on 21 April 2026 and will implement it “exclusively through” the AI Act and other EU law (Council Decision (EU) 2026/1080, art. 3).^39,41

In the United States, Executive Order 14365 (11 December 2025) told the US Attorney General to set up an AI Litigation Task Force whose “sole responsibility” is to challenge state AI laws that conflict with the order’s policy (s 3). It also called for a Federal Trade Commission policy statement (s 7); as of 23 October 2026 we found only the proposed statement (Federal Register, 7 July 2026).^14,43

Colorado repealed and re-enacted its 2024 AI law as SB 26-189 (signed 14 May 2026). From 1 January 2027 it governs automated decision-making technology used in consequential decisions, such as those on jobs, housing, lending or health care (s 5). The state Attorney General filed proposed rules on 11 August 2026 and must adopt rules by 1 January 2027.^40,44

Who is responsible for what AI creates or says?

The law looks to people and companies, as these cases show.^15,17,18

• Inventors: UK and US patent law require an inventor to be a “natural person” (Thaler v Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49, [56]; Thaler v. Vidal, 5 August 2022, p. 11).
• Copyright: the US Copyright Act requires work “to be authored in the first instance by a human being” (Thaler v. Perlmutter, 18 March 2025, p. 3), which does not bar works made “by or with the assistance of” AI where a person is the author (p. 18). The US Supreme Court declined to hear an appeal on 2 March 2026.
• Chatbots: in Moffatt v. Air Canada, 2024 BCCRT 149, British Columbia’s Civil Resolution Tribunal rejected as “a remarkable submission” the airline’s suggestion that its chatbot was responsible for its own actions. It held the airline “responsible for all the information on its website” and ordered it to pay C$812.02 in damages, interest and fees ([27], [44]).
• AI models: in Getty Images v Stability AI [2025] EWHC 2863 (Ch), the High Court held that a model which does not store or reproduce the copyright works is not an “infringing copy” ([758]); claims about training were abandoned at trial ([755]). Permission to appeal was granted ([2025] EWHC 3343 (Ch), [8], [11]).
• Legal research tools: in Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc. (29 September 2026), the US Court of Appeals for the Third Circuit held that a legal publisher’s editorial materials, copied to train a rival AI research tool, were copyrightable and that the copying was not fair use (p. 2). The appeal decided only part of the case, before trial (pp. 2, 8).^{15,16,17,18,45,46,47,48}

Can AI widen access to justice?

AI may widen access to justice, but the Legal Services Board says consumers using AI legal tools lack basic protections. It reported on 11 June 2026 that around 32% of adults in England and Wales with a legal problem get no professional support. In a survey of 1,000 adults it commissioned, 70% expected AI to make legal services easier to use. A review it commissioned found that standards written for consumer AI legal tools are almost entirely absent, and that 58% of the standards that do exist are non-binding guidance.⁴⁹

Judicial guidance in England and Wales notes that chatbots “may be the only source of advice or assistance some litigants receive” and that litigants “rarely have the skills independently to verify” the output (p. 6). The Organisation for Economic Co-operation and Development (OECD) Recommendation on Access to Justice (12 July 2023) asks for “responsible digital transformation” of justice, with trustworthy digital tools, including AI, and continued access for people who struggle with technology (para 3(a)(vii)).^19,50

Open questions on AI in law

As of 23 October 2026, we found these questions still open.

• Disclosure: whether rules settle on disclosing AI use or on certifying authorities and existing duties (see the court table).
• England and Wales: the Civil Justice Council’s final report, expected later in 2026. A June 2026 update reported broad agreement that lawyers’ drafting needs no new AI rules, with expert evidence and witness statements still under review.
• The Council of Europe Convention: when it enters into force (we found only the EU listed as a party), and for whom.
• The EU: the European Commission’s final guidelines on high-risk systems; its draft guidelines of 19 May 2026 are not yet adopted.
• US law: whether the AI Litigation Task Force challenges state AI laws in court, the Federal Trade Commission’s final policy statement, and Colorado’s final rules.
• Pending cases: the appeal in Getty Images v Stability AI, the rest of Thomson Reuters v. ROSS Intelligence, and, in Jackson Federation of Teachers v. Fitch, any rehearing and the case’s return to a new judge.^{7,11,14,23,26,39,43,44,47,48,51,52,53}