THE AILAW REPORT - FRIDAY 6TH NOVEMBER

ISSUE 013 | FRIDAY 6 NOVEMBER 2026

Combined reporting period
23 October–6 November 2026

US appeal court rejects an unsupported AI-based challenge to lawyers’ fees.

UK proposes a statutory framework for facial recognition and biometric oversight.

Meta faces a UK competition investigation and interim publisher protections in Brazil and Norway.

New York enacts restrictions on unsafe AI companion features for minors.

Google and Snap face new child-safety lawsuits involving AI services.

Louize Clark | Published by AI Policies UK
Observe. Examine. Inform.

This Fortnight

The period brought decisions about the use of AI in legal work, new disputes over access to content and data, and further steps towards governing autonomous systems. A US appeal court rejected an unsupported claim that AI assistance made lawyers’ actual hours unreasonable. Other courts examined defective AI-assisted submissions, the responsibility for checking the final document and the consequences of relying on invented authority. In South Africa, published reasons addressed an arbitrator’s reliance on a non-existent judgment.

Publisher-content disputes developed on separate tracks. The UK Competition and Markets Authority opened an investigation into Meta’s commercial terms; Brazilian and Norwegian courts reportedly granted interim relief to media organisations; and Türkiye’s competition authority described interim measures concerning Google’s AI search services. These developments have different legal effects and do not amount to a single international ruling on AI training.

The regulatory record also became more specific about agents. UK regulators issued a joint statement, Hong Kong introduced further work on agentic finance, the US OCC published an examiner framework, and Singapore issued an advisory on personal AI agents. New York’s companion legislation was enacted, while the US Treasury proposed expanded outbound-investment controls involving AI development. The Bank of England and FCA published a further survey of AI adoption and dependencies in UK financial services. Fresh claims concerning Google’s school-accessed AI and Snap’s My AI added to the child-safety litigation record; those allegations remain unadjudicated.

Coverage and source standard

This is an exceptional combined edition. It covers developments occurring or officially published from 23 October to 6 November 2026, using sources available when prepared on 6 November. Developments already reported in Issue 012 are not counted again simply because 23 October falls in both periods. Later procedural events are identified as updates. Earlier or undated decisions newly reported during this period appear separately.

The report distinguishes judgments, interim orders, complaints, investigations, enacted legislation, commencement, proposals, guidance and research. “High” describes the evidence supporting the reported event, not the likelihood that an allegation will succeed. “Qualified” identifies a material source limitation. Original records and official publications are preferred; reliable reporting is used with express attribution where the underlying document was unavailable.

Coverage is international and selective. It is not a census of every filing, local proceeding or official publication worldwide. Routine appearances, transcript requests and repeated alerts are not treated as separate substantive developments.

Contents

Courts and litigation

Regulators, enforcement and professional guidance

Government, legislation and public authorities

Earlier decisions newly reported and publication updates

Previously reported matters and dates to watch

Editorial verification note

Courts and litigation

ALR-013-01 | Appeal court rejects an unsupported AI-based challenge to lawyers’ fees

Jurisdiction: United States | Court: Tenth Circuit Court of Appeals
Matter: Flux Drive Inc and others v John Qiang Rong and others; appeals 25-1308, 26-1013 and 26-1292
Area: Costs; professional responsibility; AI-assisted research
Status/stage: Published appellate judgment | Event date: 30 October 2026 | Confidence: High

What happened. The court affirmed dismissal of claims against the Rong defendants and the associated fee orders. The appellants argued, among other things, that defence lawyers should have billed fewer hours because they used AI and online research tools. The court found no evidence that AI use had produced excessive billing. It recognised that technology could reduce necessary hours, but also that lawyers must verify authorities and legal arguments.

Underlying question. Whether AI assistance made the hours actually charged unreasonable on the evidence before the court.

Limits. The judgment does not authorise billing for work never performed or decide how every AI-assisted retainer should be priced. It upheld the assessed hours and rates in these appeals.

Verification: Published court opinion, especially the discussion of fees.

ALR-013-02 | South African reasons address an arbitrator’s reliance on invented authority

Jurisdiction: South Africa | Court: Labour Court
Matter: South African Post Office SOC Ltd (in business rescue) and others v Commissioner Menzi Fakude NO and others; case 2026-249228
Area: Arbitration; business rescue; verification of legal authority
Status/stage: Written reasons for an interim stay | Event date: Reasons 23 October; order 15 October 2026 | Confidence: High

What happened. The court explained its decision to stay unfair-dismissal arbitration pending review. The arbitrator had relied on a purported Constitutional Court decision concerning the Post Office. The cited neutral reference belonged to a different case about immigration. The court considered the apparent use of unverified AI research and the arbitrator’s failure to apply his mind to the governing law.

Underlying question. The effect on a jurisdictional ruling of reliance on a non-existent authority.

Limits. The 23 October development is publication of reasons, not a new stay made that day. The arbitration was restrained pending review; the review and underlying dismissal dispute were not finally determined by this interim order. AI generation was inferred rather than established through a technical examination.

Verification: Original judgment reproduced in the AI Hallucination Cases Database.

ALR-013-03 | Delhi court grants limited interim protection against impersonation and deepfakes

Jurisdiction: India | Court: Delhi High Court
Matter: Kriti Sanon v BBE Marketing Inc and others; CS(COMM) 1132/2026
Area: Privacy; publicity; impersonation; synthetic media
Status/stage: Ad interim arrangement | Event date: 29 October 2026 | Confidence: High

What happened. The court addressed identified uses of the actor’s identity, including explicit synthetic images, impersonating accounts, chatbots and voice models presented as the actor. It confined temporary protection to specified harmful content and identified URLs. It expressly reserved the wider legal questions about the rights asserted for consideration after the defendants were heard.

Underlying question. Which established causes of action protect a person against particular forms of AI-enabled impersonation and misuse of identity.

Limits. This was temporary relief, not final recognition of an unrestricted standalone personality right. The order preserved legitimate news, criticism, commentary, parody, satire and other expression outside its directions.

Verification: Court order reproduced by Indian Kanoon.

ALR-013-04 | Bombay customs challenge ends in a reported consent remand

Jurisdiction: India | Court: Bombay High Court
Matter: Quanta Logistics Pvt Ltd v Commissioner of Customs Import-I and others; writ petition 5093/2026
Area: Administrative adjudication; alleged AI-generated authorities
Status/stage: Consent remand reported | Event date: Final order reported as 30 October; accounts published 5–6 November 2026 | Confidence: Qualified

What happened. Contemporary legal reports describe the setting aside, by consent, of a customs adjudication order after the company challenged two purported Supreme Court authorities as non-existent and apparently AI-generated. The matter was remitted for fresh adjudication. The retrieved 27 October court order confirms that argument had concluded, remand had been raised and final orders were listed for 30 October.

Underlying question. The validity of an administrative decision challenged for reliance on unverified or invented precedent.

Limits. The final remand order was not independently retrieved. The reports describe a consent disposition, not a separate judicial finding establishing how the disputed authorities were generated. The 27 October listing order is not the final ruling.

Verification: 27 October court order; dated legal account; 6 November report.

ALR-013-05 | OpenAI challenges a British Columbia privacy production order

Jurisdiction: Canada — British Columbia | Court/authority: Supreme Court of British Columbia; OIPC
Matter: OpenAI OpCo LLC; judicial review of Order P26-16; Vancouver Registry S-267922
Area: Privacy investigation; compulsory information; chatbot processing
Status/stage: Judicial review initiated | Event date: Judicial review announced by the regulator on 27 October 2026 | Confidence: Qualified

What happened. The privacy commissioner announced that OpenAI had challenged a document-production order connected with the investigation arising from the Tumbler Ridge shooting. The official judicial-review register confirms the proceeding and records that a court date had not yet been set. The underlying September order requires information and documents under British Columbia’s Personal Information Protection Act.

Underlying question. The regulator’s authority to obtain the requested material and investigate the relevant handling of personal information.

Limits. Commencing review does not establish that the order is invalid, stayed or complied with. The petition and a judicial ruling on its merits were not independently retrieved.

Verification: OIPC judicial-review register; official order register; OIPC announcement index.

ALR-013-06 | Brazilian media associations obtain reported interim relief against Meta

Jurisdiction: Brazil — São Paulo | Court: 44th Civil Court of São Paulo
Matter: Proceedings brought by ANJ, ABERT and ANER concerning Meta’s commercial terms
Area: Publisher content; contract terms; copyright; AI training
Status/stage: Interim injunction reported | Event date: 3 November 2026 | Confidence: Qualified

What happened. The claimant associations announced an injunction suspending parts of Meta’s proposed commercial terms and restricting unauthorised collection and use of their members’ journalistic content, including for AI development and training. Their account says Meta must maintain the affected accounts and services without penalising refusal of the terms. Brazilian reporting corroborates the announcement.

Underlying question. Whether access to platform services may be conditioned on the contested permissions to collect and exploit publisher content.

Limits. The full signed order was not retrieved. This is interim relief described by the claimants and contemporaneous reporting, not a final copyright judgment or a worldwide prohibition on Meta’s AI training.

Verification: ABERT announcement; independent report.

ALR-013-07 | Norwegian media obtain reported interim relief against Meta’s commercial terms

Jurisdiction: Norway | Court: Oslo District Court
Matter: Proceedings by Norwegian media groups against Meta Platforms Ireland Ltd
Area: Competition; publisher content; contract terms; AI training
Status/stage: Interim injunction reported | Event date: 28 October 2026 | Confidence: Qualified

What happened. The Norwegian Media Businesses’ Association and claimant Amedia announced that the court had restrained implementation of Meta’s revised commercial terms pending a final determination of their legality. The publishers challenged terms they said would permit extraction of journalism, including paywalled material, for AI services. Contemporary NTB reporting corroborates the interim disposition.

Underlying question. Whether a platform may condition publishers’ continued access on the disputed content-use terms.

Limits. The full court order was not independently retrieved. This is attributed reporting of temporary relief, not a final merits judgment or a worldwide prohibition. It is separate from the Brazilian action and the UK investigation.

Verification: Norwegian media association announcement; Amedia announcement; NTB report reproduced by Rett24.

ALR-013-08 | Australian deepfake prosecution reaches sentence

Jurisdiction: Australia — South Australia | Court: District Court
Matter: Commonwealth prosecution of William Hamish Yeates
Area: Criminal law; non-consensual synthetic imagery
Status/stage: Sentence reported | Event date: 27 October 2026 | Confidence: Qualified

What happened. ABC reported that Yeates received a sentence of two years and three months following guilty pleas concerning the creation or alteration and distribution of non-consensual synthetic sexual material and related communications offences. He was released under a recognisance release order rather than required to serve that term immediately in custody.

Underlying question. The application and sentencing consequences of Commonwealth offences addressing digitally created or altered sexual material.

Limits. The sentencing remarks were not retrieved. The sentence and release mechanism are attributed to contemporaneous court reporting; the outcome does not indicate that the conduct was lawful or create a general sentencing rule.

Verification: ABC court report, 27 October.

ALR-013-09 | Axon faces a proposed class action over automated licence-plate surveillance

Jurisdiction: United States — California | Court: Northern District of California
Matter: Ryan Mun and others v Axon Enterprise Inc; 3:26-cv-13127
Area: Privacy; computer vision; surveillance data
Status/stage: Complaint filed | Reporting date: 4 November 2026; filing date differs between reproduced listings | Confidence: Qualified

What happened. Reuters reported a proposed class action alleging unlawful collection and retention of California drivers’ information through Axon’s automated licence-plate reader network. A court-feed reproduction records a complaint on 4 November and subsequent summons and scheduling activity.

Underlying question. Whether the collection, aggregation and retention alleged in the complaint violate the applicable privacy protections.

Limits. The originating complaint was not independently reviewed. The class has not been reported as certified, and allegations are not findings. Reproduced case listings differ between 3 and 4 November. This edition dates the item by the 4 November report rather than treating either filing date as independently settled.

Verification: Reuters; court-feed reproduction.

ALR-013-10 | Google faces product-liability and chatbot-law claims over school-accessed AI

Jurisdiction: United States — California | Courts: Santa Clara County Superior Court; removed federal proceeding
Matter: Emma Doe v Google LLC and Alphabet Inc; state case 26CV507724; federal case 5:26-cv-13205
Area: Product liability; warnings; children; AI search and chat
Status/stage: Complaint filed and removed | Event dates: State filing 2 November; federal filing 4 November 2026 | Confidence: High for the filed complaint

What happened. A minor, through her guardians, brought claims alleging that school-accessed Gemini-powered AI contributed to a serious psychological crisis and that the defendants failed to provide appropriate safeguards and warnings. The pleading includes design-defect, failure-to-warn, negligence, concealment and California Companion Chatbot Law claims. Google told Reuters that it was reviewing the claims and described work on youth safeguards.

Underlying question. Whether the design, provision and warnings alleged in the complaint give rise to the pleaded statutory and common-law liability.

Limits. The allegations and causal account have not been judicially determined. Removal is a procedural event. The pleading’s court stamps distinguish the 2 November state filing from its 4 November filing in the federal case.

Verification: Filed complaint supplied by claimant counsel; Reuters account including Google’s response.

ALR-013-11 | Perplexity subscription terms are challenged in a proposed class action

Jurisdiction: United States — California | Court: Northern District of California
Matter: Bonga v Perplexity AI Inc; 3:26-cv-12673
Area: Consumer protection; paid AI services; subscriptions
Status/stage: Complaint filed | Event date: 28 October 2026 | Confidence: High for the filed complaint

What happened. The public docket confirms a class-action complaint by Ntsikelelo Bonga against Perplexity. The filed complaint alleges that Perplexity failed to disclose the usage allowances attached to Pro subscriptions and reduced benefits during paid subscription periods without notice. It seeks public and class-wide relief under consumer-protection laws and pleads unjust enrichment in the alternative.

Underlying question. Whether the subscription representations and changes alleged in the action meet applicable consumer-law requirements.

Limits. The filing establishes commencement only. The pleading is an allegation by the claimant. No class certification or finding of liability is established.

Verification: Public docket; filed complaint reproduced by Truth in Advertising.

ALR-013-12 | Britannica ordered to clarify excluded artwork in Perplexity copyright case

Jurisdiction: United States — New York | Court: Southern District of New York
Matter: Encyclopaedia Britannica Inc and Merriam-Webster Inc v Perplexity AI Inc; 25-cv-7546; ECF 152
Area: Copyright; scope of asserted works; discovery
Status/stage: Discovery request granted in part | Event date: 27 October 2026 | Confidence: High

What happened. On Perplexity’s discovery request, Magistrate Judge Sarah L. Cave required the publishers to supplement their response to Interrogatory 3 by clarifying which two-dimensional artwork was excluded from the works asserted in the case. The court declined to require a page-by-page or line-by-line inventory and denied the remainder of the request. The supplemental response is due on 10 November.

Underlying question. The specificity required in identifying copyright works and excluded material during discovery in this AI dispute.

Limits. The order does not determine copying, infringement, fair use or damages. It should not be confused with a 26 October counsel-withdrawal order or the 22 October case-management directions.

Verification: Original opinion and order reproduced by Justia.

ALR-013-13 | Further compliance orders follow the Fontainebleau AI-conversation ruling

Jurisdiction: United States — Florida | Court: Southern District of Florida
Matter: Fontainebleau Florida Hotel LLC v Botach; 1:25-cv-20251
Area: AI conversations; discovery; work product
Status/stage: Production deadline and emergency-stay dispute | Event dates: 28 and 30 October 2026 | Confidence: Qualified

What happened. A source-linked litigation account describes a further order requiring responsive non-privileged AI communications and a privilege log by 30 October. It also records the district judge’s refusal that evening to suspend the deadline and the defendant’s objections to specified parts of the earlier ruling.

Underlying question. Compliance with targeted AI discovery while preserving document-specific objections.

Limits. The later orders were not independently retrieved in full. No conclusion is drawn about actual production or the disposition of later objections. The underlying 21 October decision, reported at ALR-012-03, did not make every AI conversation discoverable or eliminate work-product protection.

Verification: Dated, source-linked litigation account; original 21 October order.

ALR-013-14 | Film-school litigation produces a $5,000 sanction and training requirements

Jurisdiction: United States — California | Court: Central District of California
Matter: United States ex rel Phillips v Los Angeles Film School and others; 2:24-cv-05214; ECF 181
Area: AI-assisted submissions; Rule 11; supervision
Status/stage: Dismissal and sanctions order | Event date: 23 October 2026 | Confidence: High

What happened. The court imposed a $5,000 sanction on counsel Mark Kleiman, required notification to the California State Bar, and ordered at least eight hours of ethics training, including a course addressing generative AI. Its findings included the filing of a ChatGPT-assisted opposition prepared by the relators without adequate legal review. The claims were dismissed without leave to amend.

Underlying question. Responsibility for signing and filing material that counsel has not adequately checked, including material drafted by clients using AI.

Limits. The result rested on the record of misconduct and litigation failures as a whole. It does not prescribe a universal penalty for AI errors. The court discharged the show-cause process as to another lawyer who had not been responsible for the filings.

Verification: Original order.

ALR-013-15 | California lawyer must report non-monetary sanctions to the State Bar

Jurisdiction: United States — California | Court: Central District of California
Matter: Chechitelli v Etiwanda School District and others; 5:26-cv-03777; ECF 26
Area: Professional competence; unsupported claims; AI drafting
Status/stage: Non-monetary sanctions | Event date: 26 October 2026 | Confidence: High

What happened. The court ordered counsel to report the sanctions order to the California State Bar by 13 November and file proof by 25 November. The order records inadequate prefiling investigation and the use of Google and AI tools to draft claims without a legal basis. The court considered a financial sanction insufficient in the circumstances.

Underlying question. A lawyer’s responsibility to investigate and understand claims before putting them before the court.

Limits. A self-reporting order is not a final State Bar disciplinary determination. The court’s conclusions concern counsel’s conduct in this case.

Verification: Original sanctions order.

ALR-013-16 | Court emphasises checking the final filed document after a Legora drafting error

Jurisdiction: United States — California | Court: Northern District of California
Matter: Farhat v Cutera Inc; ECF 27
Area: Filing accuracy; delegation; generative AI
Status/stage: Show-cause order discharged; no sanction imposed | Event date: 28 October 2026 | Confidence: High

What happened. Counsel acknowledged using Legora for an initial draft and said that corrected authorities had been supplied before filing, but an earlier version was converted to PDF and filed. The court accepted the explanation and corrections for the purposes of discharging the show-cause order. It stressed counsel’s responsibility to review the final document, irrespective of AI use or delegation to staff.

Underlying question. Whether checking an intermediate draft fulfils responsibility for the document ultimately filed.

Limits. The court declined sanctions at that stage. It would disregard the challenged authorities and would not consider the replacement authorities supplied in the declaration; it did not endorse the drafting process.

Verification: Original two-page order.

ALR-013-17 | Wyoming court strikes a motion and requires explanations for suspect authorities

Jurisdiction: United States — Wyoming | Court: District of Wyoming
Matter: Clifford v Neidle and others; 2:26-cv-00235; ECF 10
Area: Suspected AI drafting; jurisdiction; service; sanctions procedure
Status/stage: Motion struck; show-cause directions | Event date: Filed 28 October 2026 | Confidence: High

What happened. The court expressed concern that submissions had been prepared using AI without adequate oversight. It struck the alternative-service motion and required production of identified authorities or sworn explanations, alongside explanations concerning other matters in the filings. Deadlines fell on 30 October and 3 November.

Underlying question. Whether the filings satisfied procedural obligations and the duty to provide authentic, legally supported submissions.

Limits. The order initiated a sanctions process; it did not itself impose a final monetary sanction. The outcome after the response deadlines was not established in the reviewed record.

Verification: Original filed order.

ALR-013-18 | New York court imposes a $250 sanction after counsel acknowledges unchecked AI use

Jurisdiction: United States — New York | Court: Supreme Court, New York County
Matter: PC-59 Doe v Jewish Board of Family and Children’s Services Inc; 2026 NY Slip Op 51818(U); index 950693/2021
Area: False quotations; professional responsibility
Status/stage: Sanction imposed | Event date: 28 October 2026 | Confidence: High

What happened. Following a 23 October direction to explain misleading legal submissions, counsel acknowledged using an AI tool and failing to verify its output. The court imposed a $250 sanction, taking account of the acknowledgement, acceptance of responsibility and the apparent absence of earlier similar conduct.

Underlying question. The appropriate response to false quotations and unsupported legal assertions in AI-assisted filings.

Limits. This was a case-specific exercise of sanctions powers. The modest amount is not a general tariff, and candour did not remove responsibility for the filing.

Verification: Original judicial text.

ALR-013-19 | Canadian Federal Court addresses apparent AI citations in a visa-review application

Jurisdiction: Canada | Court: Federal Court
Matter: Vokshi v Canada (Citizenship and Immigration); 2026 FC 1346; IMM-10807-25
Area: AI disclosure; accuracy; self-represented litigants
Status/stage: Judicial review dismissed | Event date: 3 November 2026 | Confidence: High

What happened. The court identified numerous citations whose case names and propositions did not match the authorities. It considered the pattern suggestive of AI generation and criticised failure to comply with the court’s AI declaration requirements or address the errors. It nevertheless examined the visa decision and dismissed the application without costs.

Underlying question. The responsibility to disclose relevant AI use and verify authorities, including where the record was initially prepared without a lawyer.

Limits. AI use was inferred from the citations. The court’s treatment of the immigration merits must not be reduced to a claim that using AI automatically defeats an application.

Verification: Original reasons and judgment.

ALR-013-20 | Further court decisions distinguish errors, warnings and actual sanctions

Area: Accuracy of submissions; suspected AI use | Confidence: High for the judicial texts identified below

Contreras Ramirez v Blanche — Tenth Circuit, 28 October; 25-9577. The court identified nonexistent and misrepresented authorities, suspected generative AI use and warned the self-represented petitioners. It denied review but expressly declined sanctions. Original decision.

Nabors v Nabors — Indiana Court of Appeals, 29 October; 26A-DC-699. In a custody appeal, the court admonished the appellant for a fictitious authority and warned lawyers and litigants to verify AI-generated citations. The warning does not establish which tool produced the citation. Original decision.

McCray v Lee — Illinois Appellate Court, 26 October; 2026 IL App (2d) 250559-U. The court affirmed a protection order and treated unsupported arguments as forfeited. It discussed the increasing problem of AI hallucinations but did not establish the source of the appellant’s nonexistent authorities. Original decision.

Lewis v Cameron-Nichols — Arizona Court of Appeals, 23 October. The court affirmed the family-law rulings, identified apparent AI-related briefing problems and awarded a reasonable portion of the respondent’s appellate fees and costs. It did not specify a fixed dollar award in the decision. Original decision.

Paananen v Peckham Inc — Western District of Michigan, 23 October; 1:25-cv-00573, ECF 67. The court attributed a fictitious quotation to assumed AI use, warned the self-represented litigant and declined sanctions at that time. Original decision.

Rivers v Town of Hudson — District of Colorado, 27 October; 26-cv-02636. The court required the litigant to provide the source of a quotation or explain its inclusion by 10 November. It expressly left open whether the error arose from AI or another cause. A direction to explain is not a final sanction. Original order.

ALR-013-21 | US prosecutors announce charges in an alleged AI-assisted student-aid scheme

Jurisdiction: United States — federal | Authority: Department of Justice and participating US Attorney’s Offices
Matter: Operation Ghostbusters
Area: Criminal fraud; identity misuse; AI-enabled conduct
Status/stage: Indictments and seizure announced | Event date: 29 October 2026 | Confidence: High as to the announcement

What happened. The Justice Department announced charges against five defendants and a cryptocurrency seizure exceeding $750,000 in an investigation into fraudulent student-aid accounts and enrolments. Its account describes the use of AI to impersonate students in educational activity and complete coursework within the wider scheme.

Underlying question. Criminal responsibility for the alleged identity, enrolment, financial and facilitation conduct.

Limits. The charges are allegations. The announcement does not establish that each defendant personally used AI, that every part of the investigation involved AI or that any defendant has been convicted.

Verification: Justice Department announcement.

ALR-013-22 | EVOX sues ByteDance over alleged use of vehicle photographs for Seedream

Jurisdiction: United States — California | Court: Central District of California
Matter: EVOX Productions LLC v ByteDance and others; 5:26-cv-06428
Area: Copyright; image-model training; generated outputs
Status/stage: Complaint filed | Event date: 29 October 2026 | Confidence: High for filing; qualified for complaint particulars

What happened. The public docket confirms a new copyright action. A specialist account of the complaint describes allegations that ByteDance copied registered vehicle photographs without permission and used them in connection with its Seedream image generator. The pleaded claims include direct and contributory infringement.

Underlying question. Whether the acquisition and use of the photographs and the alleged generated outputs infringe the claimant’s copyrights.

Limits. The original complaint was not independently reviewed. The assertions about copying, training and output similarity are allegations; filing does not establish infringement or the availability of any defence.

Verification: Public docket; specialist complaint account.

ALR-013-23 | Copyright litigation continues through discovery and scheduling

Status/stage: Procedural updates | Confidence: Qualified — public docket reproductions; underlying orders not all reviewed

Disney and Universal v Midjourney; 2:25-cv-05275. The reproduced court feed records discovery orders on 24 and 26 October, a 26 October hearing and further discovery activity on 28 October. Those entries establish procedural activity, not its full substance or a copyright outcome. Court-feed reproduction.

UMG Recordings and others v Suno. The recorded 30 October order amended discovery deadlines, including a 20 November deadline for third-party depositions not implicated by the circumvention claim. A scheduling amendment does not determine infringement or fair use. Reproduced docket, case 1:24-cv-11611, ECF 302.

Regulators, enforcement and professional guidance

ALR-013-24 | CMA opens an investigation into Meta’s commercial terms

Jurisdiction: United Kingdom | Authority: Competition and Markets Authority
Area: Competition; publisher content; AI development
Status/stage: Competition Act investigation opened | Event date: 30 October 2026 | Confidence: High

What happened. The CMA opened an investigation into whether Meta’s commercial terms impose unfair trading conditions on businesses and publishers. The concern includes permissions to collect website content and use it for Meta’s products, including AI features. Meta had postponed the proposed implementation and then announced a pause on 28 October.

Underlying question. Whether Meta holds a dominant position in a relevant UK market and whether the contested terms constitute an abuse under Chapter II of the Competition Act 1998.

Limits. Opening an investigation is not a finding of dominance or abuse. Initial information gathering is scheduled through January 2027.

Verification: Official case record; CMA announcement.

ALR-013-25 | Türkiye describes investigation and interim measures concerning Google’s AI search

Jurisdiction: Türkiye | Authority: Competition Authority
Area: Competition; news publishers; AI Overviews and AI Mode
Status/stage: Investigation and interim measures announced | Event date: Authority account published 2 November 2026 | Confidence: High for the authority’s account

What happened. The authority described an investigation into Google’s AI search features and interim protections for publishers. Its account says publishers should be able to decide whether their content is used without losing search visibility, and describes requirements concerning transparent and non-discriminatory content-use arrangements.

Underlying question. Whether the services and contractual arrangements harm competition or exploit or exclude publishers and competing services.

Limits. Interim intervention does not determine the final infringement question. The account does not announce a general ban on AI search, and this report does not attribute an independently verified decision number or earlier decision date.

Verification: Competition Authority news service, Turkish.

ALR-013-26 | FTC compels information on personalised prices and worker compensation

Jurisdiction: United States | Authority: Federal Trade Commission
Area: Consumer data; algorithmic pricing; gig-worker compensation
Status/stage: Section 6(b) inquiry | Event date: Announced 28 October 2026 | Confidence: High for the inquiry

What happened. The FTC issued information orders to Delta, United, JetBlue, Lyft, Uber, DoorDash, Instacart, Target and Walmart. The study examines how personal information and algorithmic tools may affect the prices offered to customers and compensation offered to workers. The orders seek the relevant inputs, systems and procedures.

Underlying question. How the investigated personalisation operates and what consequences it has for consumers and workers.

Limits. A compulsory study is not an enforcement finding that personalised pricing or compensation is unlawful. Failure to provide the required information can lead to an application to compel compliance.

Verification: FTC announcement and linked orders.

ALR-013-27 | Florida sues the operator of A2E over alleged failures to prevent abusive synthetic imagery

Jurisdiction: United States — Florida | Authority: Attorney General; civil court proceedings
Matter: State of Florida v Rengstorff Consulting Ltd
Area: Consumer protection; non-consensual synthetic imagery; product safeguards
Status/stage: Civil enforcement complaint | Event date: 28 October 2026 | Confidence: Qualified

What happened. Florida brought claims under its Deceptive and Unfair Trade Practices Act against the alleged operator of A2E. The complaint is reported to allege that published prohibitions and promises concerning consent, monitoring, account termination and reporting were not supported by effective controls. The requested relief includes an injunction, equitable relief and civil penalties.

Underlying question. Whether the product’s operation and the representations about safeguards constitute the unfair, deceptive or unconscionable practices alleged by the state.

Limits. These are allegations, not findings. The original complaint was identified but access was blocked during verification; the account is qualified and does not assign an unverified case number or report the requested remedies as granted.

Verification: Source-linked complaint account; identified complaint PDF.

ALR-013-28 | West Virginia sues Snap, including over My AI safeguards and representations

Jurisdiction: United States — West Virginia | Authority/court: Attorney General; Kanawha County Circuit Court
Matter: State action against Snap Inc
Area: Children; consumer protection; chatbot safety; personal information
Status/stage: Civil enforcement lawsuit announced | Event date: 5 November 2026 | Confidence: High for the official announcement

What happened. The Attorney General announced a consumer-protection action concerning Snapchat’s design and safety representations. The AI-specific allegations include an inadequately tested launch of My AI to children and misleading statements about private conversations and location information.

Underlying question. Whether the platform’s design and representations, including those concerning its chatbot, breach the state’s consumer-protection law.

Limits. The announcement presents the state’s allegations. It is not a finding against Snap, and this report does not attribute every harm discussed in the broader platform action to My AI. The full filed complaint was not independently reviewed.

Verification: Attorney General’s 5 November announcement.

ALR-013-29 | West Virginia separately challenges TikTok’s recommendation and safety systems

Jurisdiction: United States — West Virginia | Court: Kanawha County Circuit Court
Matter: State of West Virginia ex rel McCuskey v TikTok and others; CC-20-2026-C-1310
Area: Recommender systems; children; consumer protection
Status/stage: Complaint filed | Event date: 29 October 2026 | Confidence: High

What happened. The state filed a separate action against TikTok and related entities. It alleges misleading safety representations, harmful compulsive-use design and a recommendation system that selects and promotes content to young users. The official announcement describes comparative testing of adult and child accounts.

Underlying question. Whether the alleged recommendation, engagement and safety practices violate state consumer-protection law.

Limits. The filing establishes the claims, not their truth. This is a state enforcement action, not a certified class action or a ruling against recommender systems generally. It is separate from the Snap and earlier OpenAI actions.

Verification: Filed complaint and e-filing notice; official announcement.

ALR-013-30 | OCC issues an AI examiner framework without creating new bank requirements

Jurisdiction: United States | Authority: Office of the Comptroller of the Currency
Instrument: Bulletin 2026-54, Super Intelligence: Supervisory Framework
Area: Banking; agents; operational and model risk
Status/stage: Examiner resource issued | Event date: 4 November 2026 | Confidence: High

What happened. The OCC published a framework to help examiners evaluate banks’ use of the technologies it calls “super intelligence”. It directs attention to consequences, transferred agency, autonomy and human oversight, with examination work tailored to the bank and use case.

Underlying question. How existing supervisory work should address the distinctive operation and risks of increasingly autonomous systems.

Limits. The bulletin expressly says the framework creates no new supervisory expectations, enforceable standards or prescriptive requirements for banks. It does not require a separate AI risk category or risk appetite. The terminology does not establish that the systems examined possess general superhuman intelligence.

Verification: OCC Bulletin 2026-54.

ALR-013-31 | Bank of England and FCA publish new evidence on AI adoption and dependencies

Jurisdiction: United Kingdom | Authorities: Bank of England and Financial Conduct Authority
Area: Financial supervision; third parties; operational resilience; governance
Status/stage: Joint survey findings | Event date: 26 October 2026 | Confidence: High

What happened. The survey received 196 responses; 86 per cent of respondents reported using AI. Detailed responses identified concentrated cloud and compute provision. Of 169 reported cloud/compute relationships, 37 per cent were considered both highly critical and hard to substitute; only 35 per cent of that group had tested exit plans.

Underlying question. How financial firms’ AI use, supplier dependencies and governance arrangements are developing.

Limits. These are respondents’ reports, not a sector-wide compliance audit. Most detailed results use the full survey, completed by 105 firms, and valid question-level responses. Changes in samples, definitions and questions limit comparisons with earlier surveys. The publication creates no new rules.

Verification: Original joint survey, particularly methodology and sections 2 and 4.

ALR-013-32 | FCA reports on sponsors’ use of AI

Jurisdiction: United Kingdom | Authority: Financial Conduct Authority
Instrument: Primary Market Bulletin 67
Area: Sponsor services; due diligence; human judgement
Status/stage: Supervisory review findings | Event date: 4 November 2026 | Confidence: High

What happened. The FCA reported anonymised observations from a small sample of sponsors. Uses included research, summarisation, drafting and due diligence. Firms described human review of outputs and continuing responsibility within the sponsor team. The review also discussed enterprise tools, restrictions on consumer applications and differing governance approaches.

Underlying question. How sponsor firms retain professional responsibility and effective controls as AI becomes embedded in their work.

Limits. The small sample is not a market-wide compliance assessment. No sponsor was reported as using fully automated processes. The separate Knowledge Base consultation in the same bulletin should not be presented as new AI rules.

Verification: FCA Primary Market Bulletin 67, section 2.

ALR-013-33 | Pensions Regulator sets out how existing duties apply to AI

Jurisdiction: United Kingdom | Authority: The Pensions Regulator
Instrument: Use of artificial intelligence in pension schemes
Area: Trustee governance; service providers; member outcomes
Status/stage: Guidance published | Event date: 29 October 2026 | Confidence: High

What happened. The regulator published pensions-specific AI guidance for trustees and scheme managers. It explains the continuing responsibility for AI use, the need to understand and challenge it, and considerations including governance, data, fairness, transparency and disruption. The guidance is relevant to AI supplied through advisers and service providers as well as tools adopted directly.

Underlying question. How trustees discharge existing responsibilities when AI supports scheme administration, communication and decisions.

Limits. The guidance expressly does not establish new, separate regulatory requirements. It is distinct from the regulator’s 22 October publication about concentration risk in investment portfolios, reported at ALR-012-43.

Verification: Official guidance.

ALR-013-34 | Four UK regulators issue a joint statement on personal AI agents

Jurisdiction: United Kingdom | Authorities: CMA, FCA, ICO and Ofcom through the DRCF
Area: Agentic AI; consumer protection; data protection; regulatory coordination
Status/stage: Joint statement | Event date: 30 October 2026 | Confidence: High

What happened. The regulators published a statement on protecting people and businesses in an agentic economy. It describes work already underway, the application of existing legal frameworks and continued collaboration as personal AI agents become more available.

Underlying question. How responsibilities across competition, finance, information rights and communications regulation apply to agents acting for users.

Limits. This is a coordination statement, not a new cross-sector AI statute, authorisation scheme or assurance that every agent complies with existing law. The CMA’s associated webpage was first published in March; the October development is the joint statement.

Verification: DRCF statement publication.

ALR-013-35 | Hong Kong launches Project SAGE for agentic finance

Jurisdiction: Hong Kong | Authority: Hong Kong Monetary Authority
Area: Delegated authority; agent identity; payments; liability
Status/stage: Programme and expert group launched | Event date: 2 November 2026 | Confidence: High

What happened. Project SAGE will support exploration of agentic-finance use cases, agent identity and delegated authority, interoperability, and legal and regulatory questions. The HKMA established a cross-sector expert group to support that work.

Underlying question. How financial agents can act within verifiable authority and suitable governance arrangements.

Limits. Launching a programme does not approve individual products, settle liability questions or confer regulatory exemptions.

Verification: HKMA announcement.

ALR-013-36 | HKMA publishes a discussion paper on banks’ agentic AI governance

Jurisdiction: Hong Kong | Authority: Hong Kong Monetary Authority
Area: Banking; supervisory governance; autonomous systems
Status/stage: Discussion paper published; views invited | Event date: 2 November 2026 | Confidence: High

What happened. The HKMA published a discussion paper on banks’ use of agentic AI. It presents preliminary governance proposals and invites industry feedback, addressing accountability, traceability, human intervention, operating boundaries and arrangements for material deployments.

Underlying question. The supervisory arrangements needed when AI moves from assisting staff to taking actions under delegated authority.

Limits. Section 2.3 expressly states that the paper does not consult on new rules, state supervisory expectations, create new obligations, or modify or interpret existing obligations. It is a discussion document, and the earlier 15 October guidance is not newly issued in November.

Verification: HKMA official announcement; full discussion paper, especially section 2.3.

ALR-013-37 | Hong Kong SFC publishes its first large-firm AI readiness survey

Jurisdiction: Hong Kong | Authority: Securities and Futures Commission
Area: Financial supervision; agentic AI; governance and controls
Status/stage: Survey and supervisory direction | Event date: 2 November 2026 | Confidence: High

What happened. The survey covered 20 large firms and 78 affiliated licensed corporations. Thirty per cent reported using agentic AI and 25 per cent piloting it. The SFC described a supervisory approach centred on safety, human responsibility and collaboration, and said the findings would inform its work.

Underlying question. How governance and control arrangements develop as AI becomes connected to internal systems and operates with greater autonomy.

Limits. These are responses from the specified sample, not adoption rates for all Hong Kong financial firms or a finding that the systems were compliant. The publication is not itself a new binding agentic-AI code.

Verification: SFC official release 26PR182.

ALR-013-38 | Canadian privacy regulators call for safeguards for AI in health

Jurisdiction: Canada | Authorities: Federal, provincial and territorial privacy commissioners and ombuds
Area: Healthcare; research; sensitive data; accountability
Status/stage: Joint resolution and recommendations | Event date: 29 October 2026 | Confidence: High

What happened. The regulators called for responsible development, procurement and use of health AI. They identified six principles: protection of privacy, validity and reliability, safety, human rights, transparency and accountability. Recommendations address governments, developers, suppliers, healthcare adopters and research ethics bodies.

Underlying question. How health systems preserve privacy and public trust while adopting AI that can affect patient care and research.

Limits. The resolution calls for legislative and organisational action; it does not itself enact those changes or determine a complaint against a particular supplier.

Verification: Official joint-resolution announcement.

ALR-013-39 | Brazil publishes further findings from its AI and data-protection sandbox

Jurisdiction: Brazil | Authority: ANPD
Area: AI testing; sensitive data; explainability; supplier governance
Status/stage: Monitoring reports 4, 5 and 6 published | Event date: 3 November 2026 | Confidence: High

What happened. ANPD published the final three partial monitoring reports from its pilot with CIAAM/USP. The authority describes further testing of mobility, health and financial solutions, including accessibility, location-data anonymisation, supplier data sharing, transparency about clinical recommendations, and explanations for automated preventive blocking.

Underlying question. Which technical and governance controls can be tested in controlled AI deployment involving personal information.

Limits. ANPD describes the sandbox as experimental and preventive, supporting institutional learning. The reports are not general product approvals, final regulations or enforcement findings against the participants.

Verification: ANPD publication notice and reports.

ALR-013-40 | Ofcom opens two investigations concerning illegal intimate imagery

Jurisdiction: United Kingdom | Authority: Ofcom
Area: Online safety; synthetic intimate imagery; platform controls
Status/stage: Investigations opened within an existing programme | Event date: 26 October 2026 | Confidence: High

What happened. Ofcom announced investigations into the providers of two adult services within its programme on non-consensual intimate imagery. The providers were not named because of the nature of the services. The programme examines compliance with illegal-content duties, including the use of hash matching or other measures.

Underlying question. Whether relevant services meet their duties to address the dissemination of illegal intimate imagery, a category that includes qualifying AI-generated deepfakes.

Limits. The programme began in September; the two investigations are the new October event. Their opening does not establish a breach or that the specific material on each service was AI-generated.

Verification: Ofcom programme record and 26 October update.

ALR-013-41 | Ofcom assesses age assurance at 16; ICO stresses data protection

Jurisdiction: United Kingdom | Authorities: Ofcom and ICO
Area: Facial age estimation; children; privacy
Status/stage: Assessment and regulatory statement | Event date: 29 October 2026 | Confidence: High

What happened. Ofcom published an assessment identifying photo ID, digital ID, banking checks and facial age estimation as methods capable of highly effective age assurance at 16. It considered effectiveness, accessibility, inclusion and circumvention. The ICO stressed that age assurance must protect personal information and information rights.

Underlying question. How age restrictions can be implemented reliably and proportionately, including through AI-based estimation.

Limits. The assessment informs the development of the proposed restrictions and detailed rules. It is not certification of every age-estimation product or a statement that any method is immune to circumvention.

Verification: Ofcom assessment announcement; ICO statement.

ALR-013-42 | India consults on responsible AI in telecoms

Jurisdiction: India | Authority: Telecom Regulatory Authority of India
Instrument: Draft Guidelines on Responsible Use of AI in Telecommunication Networks and Provision of Telecommunication Services
Area: Telecommunications; sectoral AI governance
Status/stage: Draft consultation | Event date: 30 October 2026 | Confidence: High for issue and deadline

What happened. TRAI released draft sectoral AI guidelines for comment. The official consultation record identifies 27 November as the closing date.

Underlying question. The governance arrangements appropriate to AI deployed in telecommunications networks and services.

Limits. The guidelines remain draft. The consultation landing page verifies the event; the full draft was not independently reviewed for this edition, so particular governance mechanisms are not described as operative requirements.

Verification: Official consultation record; open-consultation register.

ALR-013-43 | India’s complaint-redress amendments expressly address AI agents and human assistance

Jurisdiction: India | Authority: TRAI
Instrument: Telecom Consumers Complaint Redressal (Fourth Amendment) Regulations 2026
Area: Customer support; AI agents; human oversight; accessibility
Status/stage: Amending regulations issued | Event date: 2 November 2026 | Confidence: High for the official account

What happened. The official release describes round-the-clock complaint-centre support using human representatives, automated systems, AI agents or a combination. It states that human assistance must remain available for at least sixteen hours daily and technology-based solutions must operate with appropriate human oversight. The changes also address digital access, accessibility, appeals, reporting and regulatory review.

Underlying question. How automated customer support fits within an accountable grievance-redress mechanism.

Limits. This is separate from the 30 October draft AI guidelines. The announcement verifies issuance and the described provisions; the full notified text’s commencement arrangements were not independently examined, so no immediate compliance date is inferred.

Verification: Official government release, 2 November.

ALR-013-44 | Colorado holds its AI rulemaking hearing

Jurisdiction: United States — Colorado | Authority: Attorney General / Department of Law
Area: Consequential automated decisions; conversational AI; implementation
Status/stage: Formal rulemaking hearing completed | Event date: 26 October 2026 | Confidence: High

What happened. The Department of Law held the formal hearing on proposed rules implementing the automated decision-making and chatbot-safety legislation. Its official record says written comments received by the stated 26 October deadline will be considered in drafting final rules.

Underlying question. How the enacted statutes’ implementation and reporting requirements should be specified before their 1 January 2027 commencement.

Limits. A hearing and the end of a comment period are not adoption of final rules. This edition does not treat the draft as binding or repeat the superseded commencement position of the earlier Colorado AI legislation.

Verification: Colorado official rulemaking record.

ALR-013-45 | Poland’s AI commission begins operating and approves submission forms

Jurisdiction: Poland | Authority: Commission for the Development and Safety of Artificial Intelligence — KRiBSI
Area: AI Act supervision; complaints; individual opinions
Status/stage: First meeting and administrative implementation | Event dates: 28–29 October 2026 | Confidence: High for meeting and approved arrangements

What happened. KRiBSI held its first meeting, discussed national and EU AI supervision and approved electronic forms for applications for individual opinions and complaints. The official account announced their availability from 29 October. It also addressed further interpretative and institutional work.

Underlying question. How Poland’s AI supervisory mechanisms become accessible in practice.

Limits. The meeting and forms do not establish that a particular AI system has infringed the law or received approval. This entry verifies the operational steps and does not infer the commencement of every substantive EU AI Act obligation.

Verification: Official first-meeting record.

ALR-013-46 | Civil Justice Council recommends changes for AI in court documents

Jurisdiction: England and Wales | Body: Civil Justice Council
Area: Witness statements; experts; advocacy; professional guidance
Status/stage: Final recommendations reported; implementation pending | Publication dates: Legal reporting 30 October; MedCo confirmation 2 November 2026 | Confidence: Qualified

What happened. The final report is described as recommending declarations concerning AI’s influence on factual witness statements and expert reports, with a functional approach focused on substantive content. It recommends no equivalent general declaration for legal representatives preparing statements of case or advocacy, and proposes a practice direction bringing relevant principles together. MedCo confirms publication and directs experts to the report’s expert-evidence section.

Underlying question. Which safeguards are appropriate for different kinds of documents and evidence, rather than for a tool merely labelled AI.

Limits. These are recommendations, not amendments already made to the Civil Procedure Rules. The linked official PDF could not be retrieved during verification; the substantive account is corroborated by contemporaneous legal reporting.

Verification: MedCo publication confirmation; contemporaneous legal report; independent legal account.

ALR-013-47 | Further supervisory and professional developments

FCA wholesale-bank insights — United Kingdom, 2 November. The autumn publication reports an incident in which inaccurate AI-generated code recommendations passed human review and caused loss of service. It also records unapproved AI use causing data leaks and internal-audit weaknesses involving conditional approvals, inventories, agent risk ratings and stop controls. It links back to earlier publications, which remain earlier material. The incidents’ occurrence dates are not specified. Stage: Supervisory observations, not new rules or named-firm enforcement findings. Confidence: High. FCA publication.

Central Bank of Ireland — 2 November. The bank announced that its third Innovation Sandbox Programme will focus on “Advancing AI for Consumer Good”, with applications opening in January 2027. Stage: Programme announcement; regulatory support, not product authorisation or an exemption. Confidence: High. Official announcement.

DFSA — Dubai International Financial Centre, 30 October. The regulator and Cambridge Digital Innovation and Regulation Initiative published an insights paper on frontier and agentic AI’s implications for financial-sector cyber resilience. Stage: Research and supervisory discussion, not a rule change. Confidence: High. DFSA announcement.

CFTC — United States, 28 October. The Innovation Task Force published arrangements and an agenda for its inaugural Frontier Forum on AI and agentic finance, including autonomous trading and the application of the regulatory framework. Stage: Policy forum; the agenda does not establish adopted rules or conclusions. Confidence: High for the official programme. Official agenda.

NDPC — Nigeria, reported 25 and 28 October. The commissioner’s messages called for responsible AI use and the integration of privacy, security and accountability into development and deployment. Stage: Regulatory speeches and warnings; no identified new rule or individual penalty. Confidence: Qualified — reports by the public broadcaster; exact delivery dates not independently established. 25 October report; 28 October report.

Anatel — Brazil, reported 28–29 October. Reports describe a technical standard requiring phased facial-biometric verification for mobile customers. Stage: Adopted technical standard reported; implementation is phased rather than an immediate universal disconnection requirement. Confidence: Qualified — the original standard was not retrieved, so detailed commencement claims are not reproduced. Contemporaneous account.

DOJ / Nvidia — United States, reported 28 October. Reuters, citing people familiar with the matter, reported inquiries into the competitive effects of Nvidia’s proposed acquisition of Hugging Face within a wider examination of Nvidia’s conduct. Stage: Reported investigative questioning, not a filed merger challenge, finding of wrongdoing or completed acquisition. Confidence: Qualified — no official investigative document reviewed. Reuters account.

International charity regulators — meeting 19–20 October; communiqué published 30 October. Regulators’ joint account includes digital transformation and AI among the matters considered. Stage: Published cooperation record; the meeting itself predates this edition and the communiqué does not create binding AI requirements for charities. Confidence: High. Official communiqué.

Government, legislation and public authorities

ALR-013-48 | US Treasury proposes wider outbound-investment restrictions involving AI development

Jurisdiction: United States | Authority: Treasury Department, Office of Investment Security
Instrument: Proposed 31 CFR Parts 850 and 851; 91 FR 67396; document 2026-21786
Area: National security; AI investment; countries of concern
Status/stage: Proposed rule; comments invited | Event date: 26 October 2026 | Confidence: High

What happened. Treasury proposed regulations implementing the COINS Act. Within the regime for covered transactions involving covered foreign persons, the proposal would classify development of any system within its defined AI/SI category as prohibited technology, replacing the existing distinctions based on computing thresholds and end uses. It addresses agentic systems and specifies treatment of third-party systems and internal, non-commercial use.

Underlying question. Which US outbound investments involving AI development would fall within the national-security prohibitions.

Limits. This is a proposal, not an immediately operative ban on AI investment generally. Coverage depends on the defined persons, transactions, technologies, exceptions and exemptions. Comments are due on 7 December 2026.

Verification: Official proposed rule, particularly sections 851.226(h) and 851.232.

ALR-013-49 | New York enacts restrictions on unsafe AI companion features

Jurisdiction: United States — New York | Instrument: S.9051B / A.10379C; Chapter 368 of 2026
Area: Children; AI companions; age determination; consumer protection
Status/stage: Enacted; operative date pending | Event date: Signed 28 October 2026 | Confidence: High

What happened. The legislation adds General Business Law Article 48. It restricts specified unsafe AI companion features unless the operator establishes, using permitted methods, that the user is not a covered minor. Listed features include forms of simulated human identity or relationships, manipulative interactions and specified harmful content. The text includes defined exceptions, including certain tightly confined customer-service, technical-assistance and internal business uses.

Underlying question. Which companion features may be offered to young users and how operators determine age for those restrictions.

Limits. Enactment is distinct from commencement: the operative date is 1 January 2027. The law is not an undifferentiated prohibition on every chatbot or every use of AI by a minor.

Verification: Official bill text and action history.

ALR-013-50 | Spain’s disability-rights amendments include explicit protection concerning AI

Jurisdiction: Spain | Instrument: Law 4/2026 of 1 October; BOE-A-2026-20528
Area: Equality; accessibility; disability; automated decisions
Status/stage: In force | Event date: 23 October 2026 | Confidence: High

What happened. The law’s amendments include a new additional provision concerning equality and non-discrimination in AI systems. It addresses participation of people with disabilities in design and validation, accessible design where applicable, and protection from discriminatory effects associated with opacity, bias and inadequate human supervision, including denial of reasonable adjustments.

Underlying question. How disability rights and accessibility protections apply to automated and generative systems.

Limits. The provision operates within the amended disability-rights legislation and its scope. It is not a separate comprehensive AI licensing regime.

Verification: Official consolidated text and commencement record.

ALR-013-51 | Australian clinical decision-support exemption amendments commence

Jurisdiction: Australia | Authority: Therapeutic Goods Administration
Area: Medical-device software; clinical judgement; AI decision support
Status/stage: Amendment commenced | Event date: 1 November 2026 | Confidence: High

What happened. Amendments clarifying the clinical decision-support system exemption took effect. The TGA explains that software meeting the medical-device definition generally requires inclusion in the Australian Register of Therapeutic Goods unless exempt or excluded, and that the exemption continues to concern qualifying lower-risk decision support.

Underlying question. When clinical software falls within the specified exemption rather than the general inclusion requirement.

Limits. The TGA expressly states that the amendments clarify the exemption without changing its scope or introducing new regulatory requirements. They should not be presented as a new general AI medical-device regime or blanket exemption for AI used in healthcare.

Verification: TGA commencement explanation.

ALR-013-52 | US health IT final rule removes specified decision-support certification requirements

Jurisdiction: United States | Authority: HHS / Office of the National Coordinator for Health IT
Instrument: HTI-5 final rule; 91 FR 67916; document 2026-21943
Area: Predictive decision support; transparency; certified health IT
Status/stage: Final rule published; commencement pending | Event date: 28 October 2026 | Confidence: High

What happened. The final rule removes specified source-attribute support, access and modification requirements and intervention-risk-management requirements from the decision-support certification criterion. HHS describes a wider programme of removing and revising health IT certification requirements.

Underlying question. Which transparency and risk-management functions remain required as conditions of health IT certification.

Limits. Publication does not mean all changes were already operative in October. The rule generally takes effect on 28 December, with specified instructions effective on 1 January 2027. Removing certification requirements does not repeal other applicable anti-discrimination, privacy or clinical obligations.

Verification: Official Federal Register text, including the decision-support discussion.

ALR-013-53 | Queensland introduces an offence addressing creation of synthetic intimate material

Jurisdiction: Australia — Queensland | Authority: Queensland Parliament
Instrument: Criminal Code (Non-consensual Creation of Intimate Images and Audio Material) Amendment Bill 2026
Area: Deepfakes; image-based abuse; synthetic audio
Status/stage: Introduced and referred to committee | Event date: 27 October 2026 | Confidence: High

What happened. The bill proposes an offence covering non-consensual creation of intimate images or audio through digital technology, including material that purports to depict an identifiable person. It also addresses threats and related protective measures. The government describes a proposed maximum of three years’ imprisonment for the new creation offence.

Underlying question. Protection against the harm caused by creation itself, including where material has not been distributed.

Limits. This is a bill, not an offence already commenced. Committee submissions close at 4pm Queensland time on 9 November; briefing and hearing are scheduled for 18 November and the report for 11 December.

Verification: Parliamentary committee record; official explanatory notes.

ALR-013-54 | Australia’s digital duty-of-care bill advances in the House

Jurisdiction: Australia — Commonwealth | Authority: House of Representatives
Instrument: Online Safety Amendment (Digital Duty of Care) Bill 2026
Area: Online safety; AI chatbots; harmful content; recommender systems
Status/stage: House consideration and amendments | Event date: 28 October 2026 | Confidence: Qualified

What happened. The reproduced House proceedings record consideration in detail and government amendments. The bill proposes minimum safety standards for online services including AI chatbots, with measures concerning serious harms and user choice over algorithmically recommended feeds.

Underlying question. How a preventative duty would apply to services’ design, features and management of harmful material.

Limits. This entry verifies the House proceedings from the reproduced Hansard. It does not assert Royal Assent, commencement or final passage through both Houses. Introduction on 20 October is an earlier event.

Verification: House proceedings, 28 October; government explanation of the bill.

ALR-013-55 | UK announces a statutory framework for facial recognition and biometric oversight

Jurisdiction: England and Wales | Authority: Home Office; ministerial statement
Area: Biometrics; surveillance; proportionality; discrimination
Status/stage: Consultation response and legislative plans | Event date: 4 November 2026 | Confidence: High

What happened. The government announced a planned framework for law-enforcement use of biometrics, facial recognition and related technologies, to be taken forward through the Police Reform Bill. It proposes a statutory code, independent oversight and enforcement powers, testing and safeguards. Complementary measures and a code are planned for private-sector use in crime prevention and detection.

Underlying question. The legal basis, limits and oversight for biometric identification and associated surveillance.

Limits. This is an announced legislative programme and consultation outcome, not an enacted replacement for the existing legal framework. References to a new watchdog do not mean that it was already established on 4 November.

Verification: Home Office announcement; written ministerial statement HCWS449.

ALR-013-56 | UK explains proposed resilience measures for loss of control of AI agents

Jurisdiction: United Kingdom | Authority: Government; Parliament
Instrument: Cyber Security and Resilience (Network and Information Systems) Bill and planned secondary legislation
Area: Essential services; autonomous agents; incident preparedness
Status/stage: Legislative and policy development | Event dates: 26 October statement; factsheet updated 3 November 2026 | Confidence: High

What happened. The government’s updated factsheet explains its intention to consult on using regulation-making powers to introduce measures addressing agentic loss-of-control incidents affecting regulated entities’ networks and information systems. A 26 October statement also sets out investment in AI testing, cyber defence, vulnerability identification and public-sector hardening.

Underlying question. How the essential-services security regime can address autonomous incidents that may evade or exceed existing human intervention arrangements.

Limits. The factsheet describes planned consultation and regulation-making. It should not be reported as an already operative universal duty to report every AI incident or near miss.

Verification: Agentic AI resilience factsheet; ministerial statement HCWS412.

ALR-013-57 | Singapore considers gaps in reporting autonomous AI incidents and near misses

Jurisdiction: Singapore | Authority: Ministry of Digital Development and Information; IMDA; CSA
Area: Incident reporting; autonomous systems; critical services
Status/stage: Parliamentary answer and policy study | Event date: 4 November 2026 | Confidence: High

What happened. The ministry confirmed that existing reporting requirements for serious data, cyber and regulated-service incidents apply whether or not AI contributed. It recognised that autonomous incidents and near misses may fall outside current thresholds. IMDA and CSA are studying whether additional targeted reporting or information-sharing arrangements are needed.

Underlying question. How to learn from emerging AI incidents that do not trigger existing harm-based reporting thresholds.

Limits. The answer does not announce a new mandatory near-miss regime or a final decision to adopt one. It states that potential additions should be targeted and proportionate.

Verification: Official parliamentary response.

ALR-013-58 | Singapore issues a joint advisory on personal AI agents

Jurisdiction: Singapore | Authorities: CSA, GovTech and IMDA
Area: Delegated access; personal data; prompt injection; account security
Status/stage: Joint advisory | Event date: 6 November 2026 | Confidence: High

What happened. The authorities published an advisory on the safe use of personal AI agents. It addresses the additional exposure created when agents can access accounts, information and tools and take actions for users, including the possibility of unintended or manipulated actions.

Underlying question. How users can understand and control the access and authority granted to an agent.

Limits. This is public security advice, not legislation, product certification or a guarantee that following the advice prevents every incident.

Verification: Official advisory notice; joint advisory PDF.

ALR-013-59 | China publishes network-access requirements for terminals with preinstalled agents

Jurisdiction: China | Authority: Ministry of Industry and Information Technology
Instrument: Notice Gong Xin Ting Xin Guan [2026] No. 42
Area: Phones; tablets; wearables; agent permissions and security
Status/stage: Administrative notice published | Event date: Published 6 November; instrument dated 24 October 2026 | Confidence: Qualified

What happened. The ministry’s published notice addresses devices with preinstalled agents. Its text describes adaptation testing between agents and terminals, disclosure of compatible device and operating-system information, and the incorporation of relevant checks into network-access testing, including fraud-risk and security evaluation.

Underlying question. How agent capability, permissions and execution are assessed as part of terminal compatibility and network-access controls.

Limits. The ministry’s indexed official text supports the described provisions, but direct retrieval was restricted. This entry does not assert an unverified separate commencement date or extend the notice to every standalone AI application.

Verification: MIIT official notice.

ALR-013-60 | Bosnia and Herzegovina ratifies the Council of Europe AI Convention

Jurisdiction: Bosnia and Herzegovina / Council of Europe | Instrument: CETS No. 225
Area: Human rights; democracy; rule of law; international AI governance
Status/stage: Ratification recorded | Event date: 28 October 2026 | Confidence: High

What happened. The Council of Europe Treaty Office recorded Bosnia and Herzegovina’s ratification of the Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law.

Underlying question. Implementation of the Convention’s framework through the treaty process and domestic measures.

Limits. A state’s deposit of ratification is distinct from satisfaction of the Convention’s entry-into-force conditions. The depositary’s October notice does not by itself establish that all treaty obligations were internationally operative on that date.

Verification: Treaty Office October ratifications.

ALR-013-61 | Chile’s AI proposals take separate legislative steps

Jurisdiction: Chile | Authority: Senate and parliamentary committees
Area: AI systems; robotics; cybersecurity; agents
Status/stage: Bills under consideration | Event dates: 26 and 28 October 2026 | Confidence: High for recorded stages

What happened. The Senate record shows that the Future Challenges Committee approved the general terms of bill 15.869-19 on AI systems, robotics and related technologies on 26 October, sending it to the chamber. A separate bill, 18.820-19, was introduced on 28 October to amend the cybersecurity framework in relation to private institutions operating AI agents.

Underlying question. The proposed allocation of governance and cybersecurity responsibilities for AI systems and agents.

Limits. Committee approval in general and introduction are separate stages, neither of which establishes enactment. These two bills must not be merged into a single operative AI law.

Verification: Senate committee activity record; Senate register of projects initiated in 2026; 26 October committee agenda.

ALR-013-62 | NIST publishes findings on AI incident management

Jurisdiction: United States | Authority: National Institute of Standards and Technology
Instrument: NIST IR 8621, NIST Workshop on SI Incident Management
Area: Incident response; autonomy; evidence; distributed responsibility
Status/stage: Workshop findings published | Event date: 2 November 2026 | Confidence: High

What happened. NIST published the findings of a May workshop on incidents involving AI systems. The report distinguishes attacks targeting systems from harms arising through misuse, anomalous behaviour or severe malfunction. It discusses limitations of conventional response processes, telemetry, non-human identities, scoped containment and responsibility across suppliers and operators.

Underlying question. How organisations detect, investigate, contain and recover from incidents involving probabilistic and autonomous systems.

Limits. This is a synthesis of workshop findings intended to inform future guidance and playbooks, not a new mandatory incident-reporting standard. The publication date should not be confused with the workshop date.

Verification: NIST publication record.

ALR-013-63 | NIST updates small-business cybersecurity guidance

Jurisdiction: United States | Authority: NIST
Instrument: CSWP 50, Small Business Cybersecurity: Non-Employer Firms
Area: Small businesses; AI use; information security
Status/stage: Final voluntary guidance | Event date: 2 November 2026 | Confidence: High

What happened. The final white paper supersedes IR 7621 Revision 1 and adapts the Cybersecurity Framework to businesses with no employees other than the owner. It includes consideration of AI tools and features alongside wider cybersecurity fundamentals.

Underlying question. How very small firms can establish a proportionate security programme as their tools and data dependencies develop.

Limits. The white paper is not a new legal compliance certificate or substitute for sector-specific requirements. Its scope is practical guidance, with implementation dependent on the firm’s circumstances.

Verification: NIST final publication and version history.

ALR-013-64 | Further public-authority developments

NCSC — United Kingdom, 3 November. The agency published Artificial intelligence and cyber intrusion, assessing current threat trends and possible developments to 2028. The assessment uses probabilistic language and distinguishes plausible scenarios from predictions. Stage: Official threat assessment; not a finding that a specified future attack has occurred. Confidence: High. NCSC assessment.

UK–Australia Financial Regulatory Forum — meeting 27 October; statement 29 October. The authorities recorded continuing cooperation on AI’s operational-resilience implications and potential work on AI financial advice. Stage: Joint policy statement; not mutually recognised approval or new binding AI rules. Confidence: High. Official joint statement.

Poland — 27 October. The Ministry of Justice announced a specialist team examining the legal framework for AI tools in the judiciary and prosecution service. Stage: Legal-reform working group established; not permission for automated judicial decisions. Confidence: High. Ministry announcement.

Japan — 26 October. The Cabinet Office records the seventh meeting of the AI Strategy Expert Panel. Stage: Official policy-review meeting; the record does not establish that proposed amendments were enacted or that every subsequent draft-schedule meeting occurred. Confidence: High for the meeting. Cabinet Office AI record.

Japan — 29 October and 5 November. The Consumer Commission’s AI technology subcommittee published materials for its thirteenth and fourteenth meetings on AI and consumer issues, including expert hearings. The 5 November meeting included a report on the subcommittee’s work; its minutes were still in preparation when checked. Stage: Advisory examination, not a new consumer AI law. Confidence: High for the published meeting records. Official committee index; 5 November meeting materials.

China — 3 November. An official account of the NDRC press conference describes a commitment to strengthen AI regulation and international cooperation during the Fifteenth Five-Year Plan period. Stage: Policy direction; no new comprehensive AI law is established by that account. Confidence: High for the statement. Official government account.

Earlier decisions newly reported and publication updates

ALR-013-65 | Family Court judgment examines AI’s effect on an expert report

Jurisdiction: England and Wales | Court: Family Court at Truro
Matter: Cornwall Council v M and others; [2026] EWFC 306
Area: Expert evidence; social-work assessment; AI-generated content
Status/stage: Earlier judgment newly reported | Dates: Judgment 30 September; legal reporting 5 November 2026 | Confidence: Qualified

What happened. Legal reporting describes the court’s scrutiny of an independent social worker’s use of Microsoft Copilot to organise, summarise and generate report content. The judge concluded that she could not place significant weight on the evidence in the circumstances described.

Underlying question. Whether a court can scrutinise an expert’s own analysis and reasoning when AI has contributed to the report.

Limits. This is not a judgment made in November. The judicial record confirms its identity and date, but direct access to the full text was restricted during this review; the substantive account is attributed to the legal report. The family remains anonymised. The decision does not establish that every AI-assisted expert report is inadmissible.

Verification: Judgment record; contemporaneous legal report.

ALR-013-66 | Spanish dismissal report concerns compensation, not a prohibition on AI replacement

Jurisdiction: Spain — Castilla y León | Court: High Court of Justice, Social Chamber
Area: Employment; technological change; dismissal compensation
Status/stage: Reported appellate decision; judgment date not independently established
Publication date: 27 October 2026 | Confidence: Qualified

What happened. El HuffPost reported that a sugar-processing company replaced laboratory functions with AI. Its account says the court accepted genuine technical and organisational reasons but found the dismissal unfair because compensation used an incorrect daily salary.

Underlying question. Compliance with dismissal requirements following technological change.

Limits. The original judgment and its date remain unverified. This attributed report does not establish that AI replacement itself is unlawful; no unverified case number is assigned.

Verification: El HuffPost report, 27 October; identified Spanish judiciary judgment link. The substantive account remains source-qualified pending access to that record.

ALR-013-67 | Lady Chief Justice’s annual report describes controlled judicial adoption

Jurisdiction: England and Wales | Body: Judiciary
Area: Judicial AI use; governance; access to justice
Status/stage: Annual-report findings newly reported | Publication window: Report discussed 4–6 November 2026 | Confidence: High for the annual report

What happened. The annual report describes a move towards controlled adoption of assistive tools, including testing of transcription, summarisation and age-appropriate language, supported by governance and training. It records the demands created by AI-assisted claims from litigants in person. Judges retain responsibility for decisions, reasoning and final outputs bearing their names.

Underlying question. How the justice system adopts assistance while preserving human responsibility and managing the demands placed on courts.

Limits. The report covers October 2025–September 2026. These are not all new November deployments, and the report is not itself a new procedural rule. Its AI section does not name additional tools beyond Microsoft Copilot.

Verification: Original annual report, particularly the introduction and AI section; 4 November legal account; Law Gazette, 6 November.

Editorial verification note

This edition reconciles the available court, regulator, public-authority and gap-scan material with the previous publication, supplemented by independent searches. The legal event date is used where established. Publication dates, reasons issued after an earlier order, and dates known only from reporting are separately identified. A current webpage update or crawl date is not treated as a new legal development.

Primary evidence includes judgments and filed orders, official regulatory records, legislative text, parliamentary records and treaty-depositary notices. Some originals are accessed through recognised or specialist reproductions. A claimant’s announcement verifies what the claimant reports; it is not a substitute for the court’s complete reasoning. Where an original was not reviewed, the relevant entry identifies that limit.

References in the form ALR-013-01 identify entries in this edition. Earlier published references are retained when identifying an update or avoiding duplication; these entry references do not reassign permanent matter identifiers in the underlying register. Related cases and separate legal processes remain distinct even where they concern the same company or technology.

Suspected AI use is not reported as proven. A show-cause order, referral, investigation, consultation or complaint is not a final sanction, finding or operative rule. No inference of “no developments” is drawn from an incomplete search. International coverage remains subject to publication delays, restricted records and uneven access to local-language sources. Later publications on 6 November and subsequently located in-period records may require a dated update.

The AI Law Report is an independent publication of AI Policies UK. It reports legal and regulatory developments for information and education. It does not provide legal advice or predict the outcome of proceedings.

Next scheduled edition: Friday 13 November 2026.