AI Beat

GuideFor your jobLaw

AI for lawyers: a practical guide

Which legal AI tools do what, what bar regulators and courts now expect, and the simple checks that keep invented case law out of your filings.

A stack of thick document folders
Photo: Beatriz Pérez Moya on Unsplash

In short

  • Courts have now caught AI-invented citations or quotes in more than 2,000 decisions worldwide, and the number found each quarter rose from single digits in 2023 to more than 400 at the end of 2025.
  • Tools built on a legal database, such as CoCounsel on Westlaw or Protégé on Lexis, reduce the risk. General chatbots such as ChatGPT have no verified case-law library and should never be your source for an authority.
  • The ABA's Formal Opinion 512 applies the rules you already know: competence, confidentiality, client communication, candour to the court, supervision and reasonable fees.
  • Before you put client information into an AI tool, check where it goes. If the tool can learn from it or share it, you need the client's informed consent.
  • In England and Wales, the High Court warned in Ayinde (2025) that lawyers who cite fake cases risk contempt proceedings and referral to regulators, and that firm leaders must put real checks in place.
  • The rule that stops almost every sanction: open and read every authority you cite, in a trusted source, before you file.

Lawyers were among the first professionals to be burned by AI, and the lesson has not sunk in everywhere. AI can already save hours on first drafts, document review and research memos. It can also invent a case, a quote or a holding that sounds exactly right. This guide covers which tools do what, what the rules say, and the habits that keep you safe.

The fake-citation problem is growing

In 2023 a New York lawyer filed a brief in Mata v. Avianca citing six cases that ChatGPT had made up. The judge fined the lawyers and their firm $5,000, and the story went around the world. It was supposed to be a one-off warning. Instead, the number of cases has exploded.

Damien Charlotin, a legal researcher, keeps a public database of court decisions that deal with AI-invented citations, quotes or arguments. It listed 2,095 decisions worldwide on 28 September 2026.

Courts keep finding AI-invented citations

Court decisions dealing with hallucinated citations, quotes or arguments, by quarter. Q3 2026 runs to 28 September.

0100200300400500Q2 '23Q4 '23Q2 '24Q4 '24Q2 '25Q4 '25Q2 '26

Source: AI Hallucination Cases Database, Damien Charlotin (CC BY 4.0)

Two things stand out in the data. First, most of the cases now come from people representing themselves, but lawyers account for close to four in ten, a share that has barely fallen. Second, courts have lost patience. Early cases often ended in warnings. Today fines, costs orders and referrals to regulators are common, and apologies rarely get a lawyer off the hook.

Which tools do what

The key question for any legal AI tool is simple: where do its citations come from? Tools that search a legal database and link to the source are much safer for research than general chatbots, which generate text from patterns and can invent authorities.

Tool Made by Best for Where its law comes from
CoCounsel Legal Thomson Reuters Research, deep research reports, drafting Westlaw
Lexis+ with Protégé LexisNexis Research, drafting, document analysis Lexis content (it replaced Lexis+ AI in 2026)
Vincent Clio (through vLex) Research across many countries vLex’s global legal library
Harvey Harvey Large firms: drafting, review, due diligence, workflows The firm’s own documents plus connected sources
Legora Legora Firm-wide drafting and review, strong in Europe The firm’s own documents plus connected sources
Spellbook Spellbook Contract drafting and review inside Word Contracts, not case law
Claude for Legal Anthropic Plugins for practice-area tasks No legal library of its own; connects to Lexis, Thomson Reuters and others
ChatGPT, Copilot, Gemini OpenAI, Microsoft, Google Summaries, plain-English letters, brainstorming No verified case-law library

Even the database-backed tools make mistakes: they can misstate what a real case says, or cite a case that has been overturned. Treat every answer as a lead to check, not an authority.

Shadow AI is common. In Thomson Reuters’ 2026 Future of Professionals survey, 74% of professionals said they use AI every week, and one in three admitted using AI tools their employer hasn’t approved. If your firm has no approved tool, people will use their own, which is exactly the confidentiality risk regulators worry about.

What the ethics rules require

There is no special AI rulebook for lawyers. Regulators apply the duties you already have.

United States. The ABA’s Formal Opinion 512 (July 2024), its first on generative AI, is the reference point. In short:

  • Competence. Understand what the tool can and can’t do, and keep that knowledge up to date.
  • Confidentiality. Before entering client information, work out whether it could be seen by people outside the firm or used to train the tool. If it could, you need the client’s informed consent.
  • Communication. In some cases you should tell clients you are using AI, for example when it affects the work or the fee.
  • Candour to the court. You are responsible for every citation and statement in a filing, whoever or whatever drafted it.
  • Supervision. Partners must set policies and train the lawyers and staff who use AI.
  • Fees. You can bill for the time you spend using the tool and checking its work, but in most cases not for learning how to use it.

States add their own guidance. Florida’s Opinion 24-1 covers the same ground. California is going further: it has proposed amendments to its professional conduct rules, including an explicit duty to check AI output. Some federal courts, such as the Northern District of Texas, require you to disclose if you used generative AI in a filing, so check the local rules and your judge’s standing orders.

England and Wales. In Ayinde v London Borough of Haringey (June 2025), the Divisional Court dealt with two cases involving fictitious citations. It said lawyers must check AI research against authoritative sources, that citing fake material can lead to contempt proceedings or referral to the police and regulators, and that heads of chambers and managing partners must put practical measures in place, not just policies. The Bar Council updated its guidance after the ruling. The SRA’s advice covers firm governance, data protection and being open with clients. Judges have their own AI guidance, last updated in October 2025.

European Union. Under the AI Act, AI used by courts or in dispute resolution to research and interpret facts and law counts as high-risk. After a 2026 amendment, those obligations apply from 2 December 2027. A lawyer’s own research tool is not high-risk, but firms still have to support AI literacy among their staff.

Five habits that prevent sanctions

  1. Read every authority. Open each case, statute and quotation in Westlaw, Lexis or an official source before you file. This single step would have prevented almost every case in the database.
  2. Use the right tool for the job. Use legal research tools for authorities. Use general chatbots only for tasks that don’t need a citation.
  3. Protect client data. Use tools your firm has approved under a contract that forbids training on your data. Remove names and identifying details when you’re not sure.
  4. Keep a record. Note which tool you used and how you checked the result. It helps if a court asks, and some courts already do.
  5. Set a firm rule. Write down who may use which tools, for what, and who checks the output. After Ayinde, that is what courts expect of firm leaders.

Want the news that changes these rules? Follow the Legal section, or look up any term in our AI glossary.

Comments

What do you think? No account needed. Please be kind; comments that attack people are removed.

We store only your name and comment. See our privacy policy.

    Sources

    1. Damien Charlotin: AI Hallucination Cases Database (CC BY 4.0, updated 28 September 2026)
    2. American Bar Association: Formal Opinion 512, Generative Artificial Intelligence Tools (July 2024)
    3. American Bar Association: ABA issues first ethics guidance on a lawyer's use of AI tools
    4. Mata v. Avianca, Inc., No. 22-cv-1461 (S.D.N.Y.), docket
    5. Courts and Tribunals Judiciary: Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank (June 2025)
    6. Courts and Tribunals Judiciary: Artificial intelligence judicial guidance (October 2025)
    7. Bar Council: Updated guidance on generative AI for the Bar (November 2025)
    8. Solicitors Regulation Authority: Compliance tips for solicitors using AI
    9. State Bar of California: Proposed amendments to the Rules of Professional Conduct related to AI (2026)
    10. The Florida Bar: Ethics Opinion 24-1 (January 2024)
    11. US District Court, N.D. Texas: Judge Brantley Starr (AI disclosure rules)
    12. Thomson Reuters: CoCounsel Legal
    13. LexisNexis: Lexis+ with Protégé
    14. Harvey: Harvey raises $550M at a $15.5B valuation (September 2026)
    15. Clio: Clio completes $1B vLex acquisition (November 2025)
    16. Anthropic: Claude for Legal plugins
    17. Thomson Reuters: Future of Professionals report 2026 (June 2026)
    18. EU AI Act, Annex III: high-risk AI systems
    19. European Commission: AI Omnibus enters into force (July 2026)

    More guides