AI Agents for Lawyers

Questions

Do I Have to Tell the Court I Used AI? Disclosure Rules by Jurisdiction

Whether you must tell the court you used AI depends on the judge: 30+ US federal districts require it, Illinois does not, England has no duty. How to check.

By Dr. Niklas SchmidtUpdated 13 November 20266 min read
On this page
  1. The answer by forum, as of September 2026
  2. Why disclosure is the wrong thing to worry about
  3. How to check a judge's requirements in five minutes
  4. Certification language that does not overclaim
  5. England and Wales, Austria, Germany and arbitration
  6. Before your next filing

It depends on the court, not the country. As of April 2026 more than 30 of the 94 US federal district courts require some disclosure or certification of AI use, and over 300 individual federal judges have their own standing orders. Illinois says no disclosure is needed. England and Wales imposes no duty to tell the judge. Everywhere, you certify that every citation is real.

The answer by forum, as of September 2026

ForumMust you disclose AI use?Detail
US federal district courtsYes in 30+ of 94 districts, by local rule or certification (map, April 2026)From a one-line certification to a bar on generative AI
Individual US federal judges300+ standing orders, counted separately from districtsJudge Brantley Starr (N.D. Tex.) issued the first in May 2023; copied by at least a dozen districts
S.D. Ohio (Judge Newman), N.D. Ohio (Judge Boyko)Generative AI prohibited outrightException for legal search engines and Microsoft products, which quietly permits Copilot
Illinois state courtsNo; Supreme Court policy since 1 January 2025AI permitted, no disclosure in pleadings, user accountable
England and WalesNo general dutyJudges answer for "all material produced in their name" (guidance of 31 October 2025); Ayinde sets expectations for advocates
Austria and GermanyNo court-facing ruleÖRAK's rules require disclosure to the client, not the court
ArbitrationWhatever the procedural order saysRaise it at the first case-management conference

The district count and the judge count are different units: a district with no local rule can still hold a judge with a strict order. And the strictest orders are prohibitions with carve-outs, so the definition of "generative AI" is what you are really reading. The standing-orders page sorts them by type.

Why disclosure is the wrong thing to worry about

Damien Charlotin's database of AI hallucination decisions stood at 2,041 on 14 September 2026, 811 of them involving lawyers. Almost none punished a failure to disclose. They punished fabricated authority, and the heaviest penalties went to lawyers who denied or hid it afterwards.

The record US sanction, $110,204.38 in Couvrette v. Wisnovsky (D. Or., 12 December 2025), covered 15 non-existent cases and 8 fabricated quotations, but what drove the figure was the aftermath: the lawyer quietly refiled amended briefs instead of confessing, conduct Magistrate Judge Mark D. Clarke called "a notorious outlier in both degree and volume". In Nebraska, a brief with 57 of 63 citations defective brought an interim suspension in April 2026 after the lawyer first denied using AI. In Ayinde v Haringey [2025] EWHC 1383 (Admin) the Divisional Court found a barrister's denial "unpersuasive". And when New Mexico's highest court dealt with a murder-appeal brief containing ChatGPT-invented witness testimony, the question from the bench was not about disclosure:

"Counsel, do you watch the news? Do you listen to the radio? ... the problem with lawyers relying on AI hallucinations is an above-the-fold story every single day." — Justice C. Shannon Bacon, New Mexico Supreme Court, as reported by Reuters, 11 September 2026

The universal rule sits underneath every local one. The State Bar of California's May 2026 guidance: "No paper filed in court should contain citations that the attorney responsible for the filing has not personally read and verified." The SRA's warning notice of 17 August 2026, after 42 AI-misuse reports in a year, requires solicitors to understand a tool "well enough to review its output critically". Disclosure orders exist because that rule was broken; the citation-checking protocol is how you keep it.

How to check a judge's requirements in five minutes

  1. Search the court's local rules for "artificial intelligence" and "generative". Use the secondary map as a first pass only; courts amend faster than trackers update.
  2. Open the judge's individual practices or standing-orders page and save a dated copy to the file.
  3. Search the docket of the matter for earlier orders mentioning AI.
  4. Check the appellate court. Whiting v. City of Athens (6th Cir., 13 March 2026) imposed $15,000 per attorney plus fees and double costs; Fletcher v. Experian (5th Cir. 2026) fined a lawyer $2,500 for 16 fabricated quotations produced with vLex/CoCounsel, so a research platform is no safe harbour.
  5. Read the definitions. An order that bans "generative AI" but excepts Microsoft products leaves Copilot in Word inside the exception and a general chatbot outside it.
  6. Where nothing requires disclosure, decide once, in the firm's AI policy, whether to disclose anyway.

Certification language that does not overclaim

Most orders accept a short certification. The mistake is drafting it as a denial. This version suits the common "used or not used, verified either way" orders; adapt the brackets to the order in front of you.

CERTIFICATION REGARDING GENERATIVE ARTIFICIAL INTELLIGENCE

Counsel certifies that: (1) generative artificial intelligence tools
[were / were not] used in preparing any portion of this filing;
(2) where such tools were used, every citation to authority, every
quotation and every characterisation of the record was checked by a
licensed attorney against the original source in [Westlaw / Lexis /
the official reporter / the court file]; (3) no portion of this filing
was submitted without attorney review; and (4) counsel remains fully
responsible for the content of this filing under [Rule 11 / Local
Rule __ / the Court's Standing Order of __].

Dated: [  ]                      /s/ [Name], [Bar No.]

Three drafting rules. Never tick "were not" if Westlaw's AI-assisted research, CoCounsel, Lexis+ Protégé or Copilot contributed. If the order asks you to identify the tool, name it and the version. And keep the verification claim literally true: "checked against the original source" means someone opened the reporter, which the brief-drafting workflow builds in as a step.

England and Wales, Austria, Germany and arbitration

The judiciary's AI guidance of 31 October 2025 tells judicial office holders not to put private information into public AI tools and holds them responsible for all material in their name; it creates no duty on advocates. What binds advocates is Ayinde v Haringey, where Dame Victoria Sharp P and Johnson J set out a ladder from admonition to contempt and police referral, ordered £2,000 plus VAT in wasted costs in each case and addressed heads of chambers and managing partners directly, as Burges Salmon's note explains. The SRA warning notice of August 2026 is the regulator's answer.

On the continent the disclosure duty runs to the client. ÖRAK's four practice rules from its first KI-Tag (19 March 2026) include "disclose AI use to clients"; the CCBE's guide of 2 October 2025 says inform clients and let them object. Germany has 11 decisions in the database, France 14 and Italy 15, so the courts see the same filings without a certification regime; the Austrian and German guidance page covers the bar rules. In arbitration the tribunal's procedural orders govern, and the arbitration page suggests raising the point at the first procedural conference.

Before your next filing

  • Confirm the local rule, the judge's order and the docket; save dated copies.
  • Run every citation through the verification protocol; a certification is a promise about that step.
  • If a fabricated authority slips through, correct the record at once; every large sanction followed concealment.
  • Tell the client whatever the court requires: 78% of clients want disclosure according to Clio's 2025 Legal Trends Report, and the client-disclosure page has the wording.

Module 6 of the AI Lab for Lawyers course works through court orders, the sanctions cases and the verification routine that keeps a filing out of the database.

Frequently asked questions

Do I have to disclose AI use in US federal court?

Only where a local rule or the judge's standing order says so, and as of April 2026 that covers more than 30 of the 94 district courts plus over 300 individual judges. The requirements differ: some want a certification that AI was not used or that all output was verified by a human, and two Ohio judges prohibit generative AI outright except legal search engines and Microsoft products. Check the judge, not just the district.

Does using Westlaw AI, CoCounsel or Microsoft Copilot count as AI I must disclose?

It depends on the order's definitions. Judges Newman and Boyko in Ohio carve out legal search engines and Microsoft products, which quietly permits Copilot. Other orders define generative AI broadly enough to catch a research platform's drafting features. Read the definition before signing, and never certify that no AI was used when an AI-assisted research or drafting feature contributed to the filing.

Do courts in England and Wales require lawyers to disclose AI use?

No general duty exists. The judiciary's guidance of 31 October 2025 addresses judges, who remain responsible for all material produced in their name, and Ayinde v Haringey set out what the Divisional Court expects of advocates: verified authorities, candour when something goes wrong and supervision by heads of chambers and managing partners. The SRA's August 2026 warning notice adds that solicitors must understand a tool well enough to review its output critically.

What happens if I used AI, did not disclose it and the court finds out?

If the filing is accurate, usually nothing, because most courts sanction fabrication rather than non-disclosure. If the filing contains fake authority, the record shows that concealment multiplies the penalty. The largest US sanction, $110,204 in Couvrette v. Wisnovsky, followed quiet refiling instead of a confession, and a Nebraska lawyer who first denied using AI received an interim suspension in April 2026.

Put this into practice

AI Lab for Lawyers: a live, hands-on course for practising lawyers

4 live sessions (8 hours), recordings included, taught by Dr. Niklas Schmidt of Wolf Theiss. Rated 4.7/5 from 381 ratings; alumni from Norton Rose Fulbright, Paul Hastings, Pillsbury and Taylor Wessing.

Reserve a seat · €200 Maven money-back guarantee · certificate of completion

Regulation & Ethics

Court Standing Orders on AI: Where You Must Disclose or Certify

Court AI disclosure rules in 2026: over 30 of 94 US federal districts and 300+ judges require certification, two Ohio judges ban it, Illinois requires none.

10 min read

Regulation & Ethics

AI Hallucination Sanctions: The Cases Every Lawyer Should Know (2023–2026)

AI hallucination cases reached 2,041 in Charlotin's database by September 2026, 811 involving lawyers. Every sanction from Mata to Couvrette's $110,204, with lessons.

12 min read

Playbooks

How to Verify AI Legal Research: A Citation-Checking Protocol

Verify AI legal research with a five-check protocol: open every authority, match the quote, confirm holding and status, log it. Template, time budgets, tools.

11 min read

Regulation & Ethics

UK Judiciary, Bar Council and Law Society AI Guidance: The 2026 Picture

UK judiciary AI guidance (Oct 2025), Bar Council considerations (Nov 2025) and Law Society material share one rule: whoever signs is responsible for every word.

8 min read

Regulation & Ethics

Do You Have to Tell Clients You Use AI? Disclosure Rules and Practice

When lawyers must disclose AI use to clients under ABA 512, California, CCBE and ÖRAK rules, when consent is required, and a sample engagement-letter clause.

10 min read

Questions

Is It Ethical for Lawyers to Use AI? The Rules in the US, UK and EU

Yes, every major bar permits lawyers to use AI. It becomes unethical when output is filed unverified or client data is exposed. US, UK and EU rules in 2026.

6 min read