The Brief

What’s changing in AI

AI regulation and court decisions, plus Inferis product news. Newest first.

Updated 4 November 2026

The Brief

Inferis's regulatory newsletter, in full. Every issue, permanently archived here, exactly as sent.

The judges wrote their own AI rules. Read them as your checklist.Issue 16 · Oct 12, 2026 He used AI to fix his AI mistake. That went wrong too.Issue 15 · Sep 28, 2026 France's notaires picked their AI. The deciding factor was where it runs.Issue 14 · Sep 8, 2026 No one reviewed the decision. Now it's an €825 million problem.Issue 13 · Sep 1, 2026 AI does not transfer your responsibility. Your regulator has now put that in writing.Issue 12 · Aug 24, 2026 704 real secrets pulled from AI's 'hidden' reasoning. 64 were never shown to anyone.Issue 11 · Aug 17, 2026 Article 50 is live. Here's what actually changed today.Issue 10 · Aug 2, 2026 Ten days to the deadline. Nineteen days without the model.Issue 9 · Jul 23, 2026

The brief

AI regulation news for practitioners.

What is changing in AI regulation and enforcement, and what it means for your work with customers and deadlines.

Newsroom

21 Oct 2026

Regulation

Law Society publishes a 58-page practice note on solicitors' professional duties when using AI

The Law Society of England and Wales published a 58-page practice note on 21 October 2026 setting out how solicitors' existing professional and regulatory duties apply when AI is used in legal practice, covering legal research, document review and disclosure, drafting, client communications and early case assessment. Its central line, carried by Society president Brett Dixon, is that professional responsibility cannot be transferred to an AI system: solicitors answer for AI-assisted work to clients, the court and third parties regardless of which tools produced it, must review AI-generated material and check citations before anything leaves the firm, and should give clients appropriate information about how AI is used in their matters, with firms able to ask clients to disclose their own AI use in turn. The note also warns that relying on AI should not stop trainees and junior solicitors from developing the core skills the profession needs.

The Law Society (England and Wales)

29 Oct 2026

Courts

England's Civil Justice Council tells witnesses of fact to declare they didn't use AI to write their statements

The Civil Justice Council, chaired by Sir Geoffrey Vos, published its Final Report on the Use of AI for the Preparation of Court Documents on 29 October 2026, closing out the working group Sir Colin Birss chaired following its February 2026 interim report and a consultation that closed in April. The Council draws a line between lawyers' own filings and witness evidence: statements of case and skeleton arguments get no new AI declaration, since legal representatives already carry responsibility under their existing duties for what they file, and the report does not recommend requiring legal representatives to disclose which AI tools they used. For trial witness statements, the Council recommends amending the Civil Procedure Rules so a witness of fact must declare that their statement was not prepared using an AI tool that generated or influenced its substantive content, including by altering, embellishing, diluting or rephrasing their own evidence. The Council also flagged AI use by litigants in person as falling outside its remit but needing urgent attention from the Civil Procedure Rule Committee. None of this is yet a Civil Procedure Rule: the recommendations still require separate action by the Rule Committee before they take effect.

Civil Justice Council / Courts and Tribunals Judiciary

22 Oct 2026

Courts

Fifth Circuit strips Mississippi judge of DEI-law case over his own chambers' AI-hallucinated order

In Jackson Federation of Teachers v Fitch, No. 25-60496, a unanimous Fifth Circuit panel (Judges Smith, Haynes and Engelhardt) vacated a preliminary injunction blocking Mississippi's anti-DEI education law, House Bill 1193, and ordered the case reassigned away from U.S. District Judge Henry Wingate. The panel found that a July 2025 temporary restraining order issued from Wingate's chambers named plaintiffs who were not party to the case, cited declarations from people absent from the record, and misquoted the statute it purported to apply; Wingate has said a law clerk used the AI tool Perplexity on an early draft that was filed by mistake before the chambers' normal review was complete. Writing for the panel, Judge Smith said the district court had carelessly exercised its Article III power by issuing sloppy edicts based on falsehoods and fabrications, and found that the corrected order which replaced the TRO, and was later folded into the preliminary injunction, still cited a fabricated 1974 Fourth Circuit case. The ruling did not decide whether HB 1193 is constitutional and did not sanction Wingate personally; it ordered the case reassigned to a different district judge on remand.

U.S. Court of Appeals for the Fifth Circuit

3 Sep 2026

Courts

D.C.'s highest local court strikes Deutsche Bank's brief over four AI-fabricated case citations

In Douglas v Deutsche Bank National Trust Co., No. 24-CV-1099, the District of Columbia Court of Appeals (the District's highest local court) issued a published per curiam order striking the appellee's brief after finding that four of its cited authorities did not exist. The underlying appeal challenges a Superior Court judgment on the pleadings granted to Deutsche Bank in a foreclosure action brought by self-represented litigant Barry Douglas. After the panel (Associate Judges Easterly and Shanker, and Senior Judge Glickman) could not locate the cited cases and ordered the trustee's counsel to show cause, attorney Loishirl W. Hall acknowledged using Google's generative AI search tool to find case authority and filing it without independently verifying the cases were real. The court held that every attorney who signed the brief bears some responsibility for its contents, not only the one who drafted it, and referred the matter for possible discipline. Writing separately, Senior Judge Glickman agreed responsibility is shared but urged a narrower reading of the court's own sanctioning authority, and said filing AI-sourced citations without verification breaches a lawyer's duty to the court and the client regardless of how the underlying case comes out.

D.C. Court of Appeals

22 Jul 2026

Regulation

Australia's Tax Practitioners Board finalises guidance on AI use under the tax agent Code of Professional Conduct

The Tax Practitioners Board published TPB(GS) 55/2026, The use of Artificial Intelligence and the Code of Professional Conduct, on 22 July 2026, finalising a draft it had put out for consultation on 24 March 2026. The guidance statement explains how existing obligations under the Code of Professional Conduct in the Tax Agent Services Act 2009 apply when a registered tax practitioner uses AI, whether a standalone chatbot or AI built into existing practice software, covering competence, reasonable care, client confidentiality, record-keeping, professional judgement, and supervision and control. Its central point: a practitioner stays personally and ultimately responsible for every tax agent service provided, and AI output is not a substitute for the practitioner's own analysis of a client's circumstances. The TPB states the guidance creates no new legal obligations beyond the Tax Agent Services Act or the general law.

Tax Practitioners Board (Australia)

25 Aug 2026

Courts

Federal Court of Australia directs parties in a wage class action to test conditioned AI use

In Rogers v McDonald's Australia Ltd (AI-use) [2026] FCA 1264, Justice Lee directed the parties to an underpayment class action against McDonald's Australia Ltd and franchisee Pollburg Pty Ltd, brought with the Shop, Distributive and Allied Employees Association, to investigate and confer on whether generative AI could help collate and synthesise group-member records spread across multiple systems ahead of mediation and a ten-week trial listed for 2028. Issued under s 37P(2) of the Federal Court of Australia Act and the Court's Generative AI Practice Note, the direction does not require AI use but sets conditions for any the parties pursue: traceability to source material, testability, confidentiality, and human supervision. It is reported as the first direction of its kind under s 37P(2) specifically addressing AI use in case preparation.

Federal Court of Australia

19 Aug 2026

Courts

Fair Work Commission orders rare costs penalty over an applicant's AI-generated unfair dismissal claim

In Khan v Aldi Pty Ltd [2026] FWC 3144, Deputy President Michael Easton ordered former Aldi employee Sadnan Khan to pay $1,230 toward Aldi's costs after Khan pursued an unfair dismissal application the Commission had twice warned him, by email on 31 July 2026, had no reasonable prospects of success: he had been dismissed three days before reaching the six-month minimum employment period the Fair Work Act requires for eligibility. Easton found a sizeable portion of Khan's submissions and correspondence were, in his words, 'obviously' AI-generated, repeating the same mistaken eligibility date across successive filings even after the warnings. Khan conceded the point and discontinued at the 5 August 2026 hearing, but the Commission made the costs order anyway, calling it rare and explicitly made for deterrence, stressing that using AI to help draft submissions is not itself unreasonable, but continuing a claim he should have known was hopeless was.

Fair Work Commission

21 Aug 2026

Regulation

Japan's Ministry of Justice issues new guidance on AI legal-support services and unauthorized practice of law

Japan's Ministry of Justice published new guidance on 21 August 2026 on the relationship between AI-based legal-support services in the business field and Article 72 of the Attorney Act, Japan's prohibition on the unauthorized practice of law. Issued by the ministry's Judicial System Department and supplementing an August 2023 guideline, it sets out when services such as AI-assisted document preparation, legal-literature research, corporate restructuring and shareholder-meeting support, and whistleblowing investigations fall outside Article 72's reach, conditioned on the service not being designed for litigation or dispute resolution and on safeguards such as lawyer supervision. The ministry also announced it will convene an expert panel within the current fiscal year, aiming to finalise binding rules the following fiscal year.

Ministry of Justice (Japan)

9 Oct 2026

Courts

UK judiciary sets five core principles for courts' own use of AI, names a central steering group

The Courts and Tribunals Judiciary published a Position Statement on the Judicial use of Artificial Intelligence on 9 October 2026, issued by Sir Colin Birss, the judiciary's Lead Judge for Artificial Intelligence. It commits to a 'cautious but positive' approach built on five principles: AI use must protect the integrity of justice and the rule of law; judicial decisions remain for a judicial office holder, and AI may assist but must never determine a case's outcome or displace judicial reasoning, evidence evaluation or discretion; judicial independence, impartiality, fairness, accountability, confidentiality and security must be preserved at all times; every AI output used in judicial work must be reviewed and approved by the responsible office holder, who remains fully accountable; and adoption proceeds cautiously through pilots, evaluation, training and governance. Only AI tools cleared by a central steering group Birss chairs may be used for judicial work, with UK data residency, enterprise-grade data protection and web access disabled; the approved-tools list itself will not be published.

Courts and Tribunals Judiciary

29 Jul 2026

Regulation

Germany's AI Act implementation law takes effect, with a dedicated chamber for justice-sector AI

Germany's Gesetz zur Marktüberwachung und Innovationsförderung von künstlicher Intelligenz (KI-MIG) entered into force on 29 July 2026, naming the Bundesnetzagentur (Federal Network Agency) as the country's central market surveillance authority and single point of contact under the EU AI Act, with BaFin retaining jurisdiction over AI in regulated financial services. The law also creates an independent AI Market Surveillance Chamber within the Bundesnetzagentur, chaired by its president and two vice-presidents, specifically for high-risk AI systems used in law enforcement, border management, and the justice system and democratic processes.

Bundesnetzagentur

21 Jul 2026

Regulation

Paris Bar Council adopts model AI usage charter for law firms

The Conseil de l'Ordre, the governing council of the Paris Bar, adopted a model charter on the use of generative and agentic AI in law firms, the second part of its 'Vers un barreau souverain' digital strategy. The charter is non-binding, offered as a reference set of good practices each firm can adapt to its size and digital maturity, built around three fixed principles: professional secrecy is an absolute limit on any AI tool's use, every AI output requires human verification, and the avocat remains fully responsible for anything produced with AI assistance. It calls on firms to vet their AI providers' data handling, hosting and governance, and ties into the national RIN/RIBP rules, GDPR, and the EU AI Act.

Ordre des avocats de Paris

24 Aug 2026

Regulation

Australia's Fair Work Commission adopts binding disclosure and verification rules for AI-drafted filings

Fair Work Commission President Justice Adam Hatcher released the Commission's final Guidance Note on the use of generative AI in Commission cases, following a March 2026 exposure draft and a commissioned research report into applicants' use of AI tools. The note applies to all individual rights matters, including unfair dismissal and general protections applications, from 20 October 2026, and sets three requirements for anyone using tools such as ChatGPT, Claude, Copilot or Gemini to prepare a filing: disclosure of when and how GenAI was used, verification that the facts, evidence, legislation and case law in the document are accurate and genuinely exist, and, for witness statements and declarations, the witness's own confirmation that the content reflects their own knowledge. Professional representatives must additionally hyperlink every case cited. Non-compliant documents may be given less weight or disregarded, attract costs orders, or see the case dismissed. Hatcher has attributed a sharp, multi-year rise in the Commission's caseload principally to the spread of GenAI tools among applicants.

Fair Work Commission

30 Jul 2026

Courts

Federal Circuit and Family Court of Australia throws out unfair-dismissal claim built on an AI chatbot's fabricated evidence list

In Ba v Sterling Parts Australia Pty Ltd, self-represented applicant Mr Ba brought an unlawful-dismissal claim against Sterling Parts Australia Pty Ltd and Sterling Parts (Melbourne) Pty Ltd under the Fair Work Act 2009 (Cth), along with claims of independent-contractor misclassification and unpaid leave and superannuation. His case relied on an evidence list of 38 documents that, the Court found, he had obtained from an AI chatbot he had come to treat as a legal adviser. When the Court ordered him to produce the underlying documents, he could not produce a single one. Dismissing the claim, Judge Symons held that an AI tool is not a lawyer and owes no duty to the Court or to the person using it, and that relying on its fabricated list bordered on contempt, calling the case 'a cautionary tale about the dangers of befriending Artificial Intelligence (AI)-powered chatbots who masquerade as legal advisors.'

The Lawyer

30 Sep 2026

Courts

Arizona court vacates manslaughter sentence over AI-generated 'victim impact' video

In State v Horcasitas, the Arizona Court of Appeals vacated the 10.5-year manslaughter sentence imposed on Gabriel Paul Horcasitas and ordered a new sentencing hearing, leaving his conviction intact. At the original sentencing, the family of Christopher Pelkey, the man Horcasitas killed in a 2021 road-rage shooting, played an AI-generated video in which a digitally recreated likeness of Pelkey appeared to speak, including forgiving Horcasitas, before the sentencing judge imposed a sentence longer than prosecutors had requested and said he found the video meaningful. The Court of Appeals held the AI-generated statement was not reliable evidence of what the deceased victim would actually have said, and that the trial court could not treat it as if it were. Believed to be the first case in which a US court has had to rule on an AI-recreated statement from a deceased victim used in sentencing.

Arizona Court of Appeals

15 Jul 2026

Courts

Arizona Court of Appeals: fabricated AI citations are sanctionable regardless of intent

In the Estate of Acciavatti, the Arizona Court of Appeals, Division One, addressed AI-generated content in a legal filing for the first time, in a probate dispute between siblings Kim Dineen and Jeffrey Kotchka. Of the eight case citations in self-represented appellant Kotchka's opening brief, the court found six deficient: some misattributing holdings or facts to real cases, two citing cases that do not exist. Kotchka said he had relied on generative AI and his own internet research and admitted he had not independently verified the citations. Affirming the lower court and awarding attorney fees against Kotchka, Judge Brian Furuya wrote that submitting hallucinated citations is unjustifiable 'intentional or unintentional,' since accepting a lack-of-intent defense would tolerate reckless indifference to the accuracy of court filings.

Arizona Mirror

8 Jul 2026

Courts

Crown Prosecution Service apologises twice after AI-hallucinated case law reaches the High Court

In Tobosaru v Court of Law Craiova, Romania, the High Court's Administrative Court heard an extradition appeal in which the Crown Prosecution Service had cited two authorities, 'Vlad v Romania [2017] EWHC 2060 (Admin)' and 'Jeziorski v Poland [2019] EWHC 3885 (Admin)', that do not exist. The CPS accepted the citations likely originated from generative AI used in legal research, and said the operative failure was a reviewing lawyer's not checking them before filing. It withdrew the authorities before the hearing, so they did not affect the outcome of the appeal, but it had to apologise to the court twice: an initial explanation crediting the appellant's own counsel with spotting the fabrications had to be corrected once the CPS admitted its own junior counsel had independently flagged them internally first. The case is believed to be the first instance of a UK prosecuting authority, rather than a private firm or individual lawyer, placing AI-hallucinated case law before a court.

Law Society Gazette

11 Aug 2026

Courts

Quebec Court of Appeal orders payment after self-represented appellant cites four non-existent AI-generated judgments

In Droit de la famille, 261234, 2026 QCCA 1092, the Quebec Court of Appeal found that a self-represented appellant in a post-divorce asset-division dispute had supported his appeal with four fabricated authorities generated by ChatGPT, including two he presented as Supreme Court of Canada decisions, along with invented quotations attributed to them. The court found he had learned during a consultation with a lawyer that the cases did not exist but filed his arguments without removing them, and only admitted the citations came from ChatGPT once the panel questioned him directly at the hearing. Calling the conduct a serious and deliberate breach under article 342 of the Code of Civil Procedure that caused unnecessary work for the court and the respondent, the court ordered him to pay his ex-wife $1,000 on top of the costs of the appeal. The panel's own reasons cited Courtready's CanLII-published research on AI-hallucinated case law in Canadian courts, calling the pattern 'a serious breach that has become recurrent.'

CBC News

25 Sep 2026

Courts

Federal Court of Australia declines to treat disclosed, verified AI use as its own basis for a costs order

In Jahani v Qiu, in the matter of Ralan Property Services Pty Ltd (receivers and managers appointed) (in liq) (costs) [2026] FCA 1419, arising from an insolvency dispute over purchaser deposits across more than 30 property developments, the plaintiffs' solicitors used generative AI to summarise a document set that grew some fifteen-fold, from roughly 500 to more than 7,500 documents, then manually reviewed and corrected the output. The court vacated the final hearing and relisted it for February 2027, ordering the plaintiffs to pay the defendants' costs thrown away, because the growth in the scale of the exercise was disclosed too late for the defendants to prepare, not because AI was used. On whether the AI use itself should attract a separate costs penalty, the court held that its Generative Artificial Intelligence Practice Note gives no separate basis for an adverse costs order over AI use that is disclosed and manually verified, a rare instance of an Australian court declining to treat AI involvement in litigation as an aggravating factor on its own.

Federal Court of Australia

3 Sep 2026

Courts

Law Society of New Brunswick fines lawyer $10,000 over AI-fabricated citations, a first for the province

In its first generative-AI misconduct matter, the Law Society of New Brunswick's discipline committee fined Miramichi lawyer Denis Sawyer $10,000 plus $2,000 in costs after he did not contest a professional misconduct complaint. Sawyer had filed a provincial court brief on 29 April 2025 in which all twelve case citations were fictitious; told of the problem, he filed a second brief that still carried two further fabricated citations. Sawyer said he had probably used ChatGPT or Claude to help draft the filing and acknowledged he had not reviewed the output before filing it. The committee also ordered him to complete four and a half hours of AI-specific professional development, at his own expense.

CBC News

16 Jul 2026

Courts

Law Society of Ontario tribunal suspends lawyer who lied about using ChatGPT in a factum

In Law Society of Ontario v. Lee, the Law Society Tribunal's Hearing Division suspended Toronto sole practitioner Mary Hyun-Sook Lee's licence for six months and ordered her to pay the Law Society $10,000, the first Canadian case to discipline a lawyer, rather than merely order costs, over generative AI misuse. Acting pro bono in a family law matter, Lee filed a factum citing four cases by hyperlink; one argued against her own client's position and three did not exist. When the judge asked whether she had used ChatGPT, she denied it outright. The Law Society's investigation established she had written the factum herself under time pressure and had not told the truth about it, either to the court or, initially, to the Society. The tribunal found she had failed to serve her client, misled the court, and failed to act with honesty and integrity.

Law Times News

23 Jun 2026

Courts

New York appellate court sanctions attorney for AI-fabricated case law in a sidewalk-injury appeal

In Landberg v City of New York, the Appellate Division of the New York Supreme Court, Second Department, sanctioned attorney Michael Sanders $8,000 and his firm, the Law Offices of Michael S. Lamonsoff, $2,500, payable to the state's Lawyers' Fund for Client Protection, after his appellate brief (prepared with generative AI) cited nonexistent cases and fabricated quotations wrongly attributed to the Court of Appeals, and mischaracterized what real cases actually held. At oral argument the panel had offered Sanders a recess to explain the citations; the court found he still failed to candidly acknowledge that the material had been AI-hallucinated. The panel separately noted, without sanctioning them, that opposing counsel had never flagged the fabricated authorities either.

New York State Law Reporting Bureau

30 Sep 2026

Regulation

California enacts first US state statute governing attorneys' and arbitrators' use of generative AI

Governor Gavin Newsom signed SB 574, the first state law in the United States to directly regulate how licensed attorneys, arbitrators, judicial officers, and certified alternative dispute resolution providers may use generative AI. Authored by Senator Thomas Umberg and passed unanimously by the Legislature on 31 August, the statute bars delegating the practice of law (or an arbitrator's decision-making) to generative AI, requires attorneys to take reasonable steps to verify AI-generated output (including every citation) and correct fabricated material before it is filed, requires disclosure of AI use to the court, and prohibits entering confidential client information into public generative AI systems. It takes effect 1 January 2027.

Office of Governor Gavin Newsom

21 Sep 2026

Courts

Federal Circuit and Family Court of Australia orders immigration lawyer to personally pay AI-fabrication costs

In Dey v Minister for Immigration and Citizenship (No 2), the Federal Circuit and Family Court of Australia dismissed a judicial review application and ordered Gurdial Singh Ranji Singh, principal of Quantum Legal Advisory & Migration Consultants, to personally pay the Minister's costs of $9,600 after the lawyer acknowledged that generative AI had produced eleven fabricated or unsupported authorities across his filings: six in the substantive judicial review submissions and a further five in his first attempt at costs submissions. The applicant was separately ordered to pay the Minister's costs of the judicial review itself, fixed at $6,700. The court described it as unacceptable that its resources had to be spent fact-checking a lawyer's citations.

The Lawyer

15 Sep 2026

Courts

Supreme Court of Tasmania voids parole gag order built on AI-hallucinated case law

In a challenge brought by convicted killer Susan Neill-Fraser, Justice Stephen Estcourt of the Supreme Court of Tasmania struck down a parole condition barring her from speaking to media about her conviction, after finding that the rationale document behind it (prepared by Parole Board secretary Ben Hancock using AI) cited case law that does not exist. The court held that the board's reliance on the fabricated authorities denied Neill-Fraser procedural fairness and made the condition legally unreasonable. Tasmania's Attorney-General demanded an explanation of the board's 'professional standards', and the state's Justice Department announced an independent review of every past Parole Board decision prepared with AI assistance, not only Neill-Fraser's.

ABC News Australia

12 Jun 2026

Courts

Ontario Law Society Tribunal orders record $31,150 costs over AI-fabricated citations

In Mazaheri v. Law Society of Ontario, the Law Society Tribunal ordered suspended lawyer Shahryar Mazaheri to pay $31,150 in costs after motion materials seeking to lift his interlocutory licence suspension cited cases that do not exist and real cases for propositions they did not support. Mazaheri admitted he had used Grok to prepare the submissions and had not checked its output before filing. Access-to-justice tracker Courtready called it the largest costs order issued by any Canadian court or tribunal to date over AI-fabricated case references.

Law Times

3 Jun 2026

Courts

Ninth Circuit suspends two attorneys, mandates AI-disclosure certification after fabricated citations

In Lnu v. Blanche, the Ninth Circuit suspended Orange County immigration attorneys Mike Singh Sethi and William Rounds from practicing before the court for six months and fined each $2,500, after their briefs cited two nonexistent cases ('Eduardo v. Garland' and 'Lay v. Holder') and misquoted several real ones. The panel said the sanction turned not on using generative AI itself but on the attorneys' lack of candor, having recast the fabrications as typographical errors rather than admitting their AI origin. For two years, filings by Sethi, Rounds, and every attorney at their firm before the Ninth Circuit must include a sworn statement disclosing any generative-AI use, naming the tool, and certifying that the signer personally verified every citation and quotation.

U.S. Court of Appeals for the Ninth Circuit

17 Sep 2026

Courts

Employment Appeal Tribunal: AI-verification duty extends to litigants in person

The Employment Appeal Tribunal held that the personal-responsibility principle requiring careful verification of AI-generated material applies to litigants in person, not only to represented parties and their lawyers. Terence Hancox, appealing the strike-out of claims against the National Farmers' Union and four of its employees, filed a skeleton argument running to roughly 300 pages and 132,000 words that he confirmed had been generated using ChatGPT; it did not comply with the EAT's Practice Direction, which calls for skeleton arguments of five to fifteen pages. Marcus Pilgerstorfer KC, sitting as a deputy High Court judge in the EAT, held that 'the underlying principles of personal responsibility and accuracy apply to all persons who submit documents to a court or tribunal,' and that 'simply submitting the product of generative AI to a court or tribunal, or placing the onus to check a document onto an opponent, is not acceptable.'

Employment Appeal Tribunal

15 Sep 2026

Courts

Los Angeles judge fines State Farm defense counsel over AI-fabricated case citations

Los Angeles Superior Court Judge Elizabeth Bradley ordered Musick, Peeler & Garrett attorney Jacquelene Robinson to pay $999.99 (one cent under the threshold that would trigger a mandatory referral to the State Bar) after filings in a Carson, California fire and storm damage claim against State Farm cited seven cases that do not exist. Robinson had used the legal AI tool Irys, believing it drew on her firm's Westlaw subscription, and did not independently check the citations before filing; co-counsel Kenneth Katel apologised to the court and to opposing counsel once the fabrications were caught.

Law360

18 Aug 2026

Courts

Washington federal court excludes expert after $3,000 sanction over AI-fabricated citations

In LeDoux v. Outliers, Inc., Judge Tiffany M. Cartwright (W.D. Wash., Tacoma) fined plaintiff's counsel $3,000 in July under Federal Rule of Civil Procedure 11 after finding dozens of inaccurate, AI-generated citations across at least five filings, the product of ChatGPT and Claude used to build a citation table that was also handed to two retained experts. On 18 August the court went further, excluding one expert's report entirely after the same fabricated citations undermined its reliability, a ruling that could cost the plaintiff her ability to prove the case.

Washington State Bar Association (NWSidebar)

9 Sep 2026

Courts

New Mexico Supreme Court holds attorney in contempt over ChatGPT-fabricated murder-appeal brief

In a 9 September order, the New Mexico Supreme Court held Santa Fe attorney Stephen Aarons in direct contempt and fined him $5,000 after he filed a murder-appeal brief containing testimony attributed to four wholly fabricated witnesses, generated by ChatGPT and never checked before filing. Aarons had used the tool to summarise the trial record for client Oscar Renee Sandoval's appeal of a first-degree murder conviction; the court struck every brief he had submitted, barred him from appearing before it pending a referral to the state's attorney disciplinary board, and appointed a public defender to refile the appeal.

The Detroit News

30 Jul 2026

Regulation

Ireland's AI Office stands up, first CEO appointed

The Regulation of Artificial Intelligence Act 2026, signed into law by President Connolly on 21 July, established the AI Office of Ireland as an independent statutory body and Ireland's central coordinating authority for the EU AI Act. The Department of Enterprise, Trade and Employment confirmed Paul Byrne, formerly of the Medical Council of Ireland, as the Office's first Chief Executive Officer; it became operational on 2 August, coordinating Ireland's existing sectoral regulators under the AI Act's distributed-authority model.

Department of Enterprise, Trade and Employment

3 Sep 2026

Courts

Solicitors Disciplinary Tribunal's first AI-hallucination case ends in strike-off

In its first case dealing with a lawyer's use of generative AI, the Solicitors Disciplinary Tribunal struck registered foreign lawyer Abhishek Kumar off the register (case 12884-2026). Kumar used AI to draft his written answer to separate SRA disciplinary charges; that answer contained misleading quotations and non-existent case citations. After the regulator flagged the errors, Kumar used AI again to prepare his explanation, which itself contained further fabricated material. The Tribunal found the harm from the AI misconduct alone was high, and noted Kumar would have been struck off for either the original conduct or the AI misuse independently.

Solicitors Disciplinary Tribunal

7 Jul 2026

Regulation

UK Jurisdiction Taskforce issues final statement on AI liability under English law

The UK Jurisdiction Taskforce published its final Legal Statement on Liability for AI Harms under the private law of England and Wales, following a public consultation opened in January 2026. The statement concludes that existing English private law resolves most AI liability questions without new AI-specific legislation: professionals can be held liable in negligence both for over-relying on AI output and for failing to use available AI tools, while developers of foundation models are unlikely to bear liability for unforeseeable downstream harms. It was prepared by the Taskforce's AI drafting team, chaired by Matthew Lavy KC, and launched at an event hosted by LawtechUK and the Law Society.

LawtechUK / UK Jurisdiction Taskforce

4 Sep 2026

Courts

Quebec's courts adopt joint guidelines on judges' use of generative AI

The Court of Appeal, Superior Court, Court of Quebec and the province's municipal courts published common guidelines, signed by all four chief judges, stating that judging is an exclusively human act. Generative AI may be used only for limited, fully human-supervised tasks such as revising a judge's own drafted text or locating information in specified sources, never for legal reasoning, assessing evidence or credibility, or drafting the substantive parts of a decision.

CBC News

20 Aug 2026

Regulation

Australian Parliament appoints a Joint Select Committee on AI

Both houses resolved to establish a Joint Select Committee on Artificial Intelligence to examine AI's risks and opportunities and review whether existing laws, including copyright, consumer protection and national security frameworks, remain adequate. The 12-member committee is accepting submissions until 14 September and must report by 30 November 2026.

Parliament of Australia

3 Aug 2026

Regulation

Legal services named first sector for the UK's Advisory AI Growth Lab

The Ministry of Justice and the Department for Science, Innovation and Technology opened applications for the Advisory AI Growth Lab, a government regulatory sandbox, with legal services chosen as its first sector. The Solicitors Regulation Authority, Legal Services Board, Council for Licensed Conveyancers and Information Commissioner's Office are all involved, giving around a dozen selected AI tools direct regulatory input while they are tested. Applications close 27 September 2026, with decisions expected by the end of October.

GOV.UK

22 May 2026

Courts

High Court rebukes Pinsent Masons over an AI-hallucinated insolvency rule

In Cork & Anor v Smith [2026] EWHC 1199 (Ch), ICC Judge Mullen found that Pinsent Masons misled the court twice over a routine block transfer application: first by citing a provision of the Insolvency Rules 2016 that does not exist, then by using generative AI again to draft an explanatory letter that repeated the error. The firm self-referred to the Solicitors Regulation Authority; the judge ordered a copy of the judgment sent to the regulator regardless.

ICLR

1 Sep 2026

Courts

Irish High Court requires GenAI declarations in court documents

Practice Direction HC 142, issued by the President of the High Court on 29 July, came into operation on 1 September. It covers court documents in civil proceedings, including pleadings, submissions, affidavits, witness statements and expert reports, and applies to proceedings whether instituted before or after that date. Affidavits and witness statements must now carry a declaration that generative AI was not used to generate their substantive content, including by altering, embellishing, strengthening, diluting or rephrasing the evidence. Expert reports must carry an equivalent declaration and disclose the nature and extent of any generative AI used in preparing them. Non-compliance can draw adverse costs orders, orders rejecting documents in whole or in part, claims or defences struck out, and referral of practitioners to the Legal Services Regulatory Authority.

Courts Service of Ireland

28 Jul 2026

Courts

Illinois appeal court raises its sanction rate for AI citations

The Illinois Appellate Court, First District imposed a $15,000 sanction on an attorney whose response brief contained false case citations and false statutory quotations produced with a commercial AI subscription, set at $1,500 for each one and deliberately above the rate other courts have imposed for similar violations. The court said it had no choice but to increase fines for AI-hallucinated citations until those fines carry a significant deterrent effect.

ABA Journal

17 Aug 2026

Product

Organisations: multiple firms under one umbrella

Practices that run more than one firm can now group them under a single Organisation, with an organisation administrator who can grant and revoke people's access across those firms from a single console. Each firm keeps its own documents, users and access rules.

17 Aug 2026

Regulation

SRA issues a formal warning notice on AI misuse

The Solicitors Regulation Authority published a warning notice covering hallucinated citations in court filings, inadequate supervision of AI-assisted work, and confidentiality risks from entering client information into AI tools. Its central line is that using AI does not reduce, remove or replace the professional, regulatory or legal responsibility a solicitor holds for the quality and accuracy of their work. The notice states that failing to have proper regard to it puts solicitors at risk of disciplinary action.

SRA

2 Aug 2026

Regulation

AI Act Article 50 transparency duties take effect

The EU AI Act's Article 50 obligations became enforceable today. Providers of AI systems that interact with people must make clear they are AI, and providers of systems that generate synthetic content must mark it in a machine-readable way (a grace period runs to 2 December 2026 for older systems). Deployers have separate duties for deepfakes and some published text.

Cloud Security Alliance

2 Aug 2026

Regulation

Luxembourg designates the CNPD as its lead AI Act authority

Luxembourg's implementing bill names the CNPD as the default AI Act market surveillance authority, with ILNAS, the CSSF, the CAA and the ILR handling specific sectors. The CNPD is required to launch its first regulatory sandbox today.

Paperjam

23 Jul 2026

Regulation

AI Act transparency deadline is ten days out

The EU AI Act's Article 50 transparency obligations apply from 2 August 2026. Providers of AI systems that interact with people must make clear they are AI, and providers of systems that generate synthetic content must mark it in a machine-readable way (a grace period runs to 2 December 2026 for older systems). Deployers have separate duties for deepfakes and some published text.

Sidley Data Matters

15 Jul 2026

Regulation

Australia announces a national Office of AI

Prime Minister Anthony Albanese announced a new Office of AI within the Department of the Prime Minister and Cabinet and a forthcoming set of Australian Standards for AI, with legislation expected in early 2027.

Prime Minister of Australia

8 Jul 2026

Regulation

EDPB adopts guidelines on anonymisation and AI web scraping

The European Data Protection Board adopted guidelines stating that GDPR governs web scraping whenever personal data is involved, and that public visibility online is not consent. Open for public consultation until 30 October 2026.

EDPB

1 Jul 2026

Regulation

APRA's CPS 230 amendments take effect in Australia

Updated operational-risk rules from the Australian Prudential Regulation Authority now treat AI vendors used by regulated institutions as material service providers, requiring a formal register and a documented fallback plan.

APRA

29 Jun 2026

Regulation

Council approves AI Act's Digital Omnibus, defers high-risk deadlines

Parliament and Council finalised amendments deferring obligations for high-risk AI systems: standalone systems now have until 2 December 2027, systems embedded in already-regulated products until 2 August 2028. Article 50's transparency deadline is unaffected.

Council of the EU

10 Jun 2026

Regulation

EU Commission publishes AI content labelling Code of Practice

The Commission's final Code of Practice on marking and labelling AI-generated content gives providers a voluntary compliance path for Article 50, drafted by six independent experts with input from more than 180 stakeholders.

European Commission

3 Jun 2026

Regulation

Cloud and AI Development Act proposed

The European Commission proposed a four-tier sovereignty framework for public-sector cloud and AI procurement. Still a proposal moving through Parliament and Council, not yet adopted.

European Commission

31 May 2026

Product

Agent Core ships

Inferis replaces its query router with a grounded, tool-calling agent that decides how to answer each question and which tools to use.

Apr 2026

Courts

Nebraska Supreme Court suspends attorney over AI citations

57 of 63 citations in a filed appeal proved defective, 20 of them entirely invented. Asked directly whether he had used AI, the attorney first denied it, then admitted it. Suspended indefinitely pending a disciplinary investigation.

WOWT

Mar 2026

Courts

Sixth Circuit sanctions attorneys over fabricated citations

Two attorneys were sanctioned $15,000 each in Whiting v. City of Athens, on top of the other side's fees and double costs, for a brief containing more than two dozen fabricated citations.

Reason, The Volokh Conspiracy

17 Mar 2026

Regulation

French Bar guidance on generative AI now in force

The Conseil National des Barreaux published a deontological guide covering professional secrecy and GDPR obligations for generative AI use by avocats. It has applied since 17 March 2026.

Conseil National des Barreaux

9 Feb 2026

Regulation

SRA restates AI duties for solicitors

The Solicitors Regulation Authority updated its compliance guidance on AI. No new rules; a restatement of existing duties under the Standards and Regulations, with one headline expectation: it must always be clear to clients when they're dealing with AI.

SRA

Jan 2026

Regulation

Seven UK tax and accounting bodies issue joint AI guidance

The bodies behind PCRT, including ICAEW and CIOT, said putting a client's data into a public AI model without consent is likely a breach of confidentiality, unless the data is properly anonymised first.

Tax Adviser

Standing

Regulation

Law Society guidance on generative AI

The Law Society advises solicitors not to put confidential data into a generative AI tool where there is no real relationship with the vendor. The guidance is part of its Generative AI essentials.

The Law Society

Background

AI use in professional firms is ahead of the policies meant to govern it. Survey figures behind these stories:

54%: of firms have no AI policy
69%: already use AI daily
85%: of clients expect AI use to be disclosed
46%: of US accounting firms have put client data into public AI

About Inferis

Inferis is document AI for businesses that can't put confidential data into public tools. Documents stay in the region you choose. Answers about documents cite the passages they draw on, and each AI answer is written to a hash-chained audit log.

Press contact

Inferis

hello@inferis.ai