The AI Law Report Issue 005 - Friday 4 September 2026
THE AI LAW REPORT
Global developments from the courts, regulators and public authorities
Issue 005 | Friday 4 September 2026Reporting period: Friday 28 August – Friday 4 September 2026Published by AI Policies UK
IN THIS ISSUE
Issue 005 records judicial, regulatory, legislative and public-authority developments identified during the reporting period from 28 August to 4 September 2026.
The court record includes eight verified judicial or court-administration developments recovered through a separate backfill after the routine Courts & Litigation discovery stream failed during part of the week. They include three newly filed US actions involving Anthropic and Suno; a US Government intervention in consolidated OpenAI copyright litigation; a published appellate order concerning fictitious AI-assisted authorities; appellate argument over an AI-affected judicial ruling; a Brazilian electoral decision defining deepfakes for the 2026 election; and the commencement of new Irish court rules governing generative AI.
The wider record includes formal regulatory action involving ChatGPT under the EU Digital Services Act, an international financial-stability intervention concerning frontier AI, large-scale AI-content enforcement in China, implementation of the Texas Responsible Artificial Intelligence Governance Act, AI-related legislative activity in California and at US federal level, new privacy proposals in Australia, AI procurement in the United Kingdom, and a range of public-sector AI and infrastructure programmes across Europe, Asia, the Middle East, Africa and the Americas.
Entries are distinguished as new matters, new events on existing matters, backfills, or partially verified developments.
The permanent ALR matter reference remains the identifier for the underlying canonical matter in The AI Law Report database. New ALR numbers for this edition should be assigned only after reconciliation against the existing register. A later order, hearing or filing in an existing matter should be added as a new event under the same ALR matter number rather than creating a new matter.
THIS WEEK
Four developments stand out from this week’s record.
The first is that AI copyright litigation is becoming more granular. The legal argument is no longer simply whether copyrighted works may be used to train models. The new proceedings against Anthropic and Suno separate questions about how source material was acquired, what happened during training, whether outputs reproduce protected expression, whether intermediaries involved in collecting data may bear responsibility, and whether rights in a performer’s identity exist independently of copyright.
Second, courts are increasingly confronting AI inside the administration of justice itself. Earlier cases largely concerned lawyers filing hallucinated authorities. This week, the District of Columbia Court of Appeals struck an appellate brief after four nonexistent authorities entered the filing, while the Fifth Circuit heard an appeal in which AI use by a judicial clerk became part of an argument for reassignment of the district judge. Ireland’s new Practice Directions take a prospective approach: AI may be used, but responsibility for what reaches the court remains human.
Third, synthetic media is moving from general policy debate into specific legal tests. Brazil’s Superior Electoral Court has now adopted criteria for identifying deepfakes during the 2026 election while separately deciding when the use of such material falls within prohibited electoral propaganda.
Finally, governments are moving beyond general AI strategies into the infrastructure and institutions through which AI will operate. The European Commission has placed ChatGPT within the enhanced supervisory framework of the Digital Services Act. Canada has set national principles for AI-linked data-centre development. South Korea is developing a specific legal framework for AI data centres. The UAE is placing specialised AI advisers inside Cabinet processes. The UK has opened sovereign AI procurement competitions for real-world government deployment.
These developments do not amount to one settled body of “AI law”. They show existing areas of law — copyright, professional responsibility, election law, privacy, employment, consumer protection, public administration and infrastructure regulation — being applied to AI at different points in its development and use.
COURTS & LITIGATION
Music publishers sue Anthropic over alleged acquisition and use of copyrighted works
Status: New canonical matter — ALR number to be assigned after register reconciliation
Jurisdiction: United States — federal
Court: US District Court for the Northern District of California
Case: Sony Music Publishing (US) LLC and others v Anthropic PBC, Dario Amodei and Benjamin Mann
Case number: 5:26-cv-09217
Area: Copyright · AI training data · Data acquisition · Generative AI
Stage: Complaint filed
Event date: 28 August 2026
Confidence: High
Sony Music Publishing, Warner Chappell and numerous affiliated music publishers have opened a new copyright action against Anthropic and co-founders Dario Amodei and Benjamin Mann.
The federal docket confirms that the complaint was filed on 28 August in the Northern District of California and identifies Anthropic, Amodei and Mann as defendants.
The publishers allege that protected musical compositions were obtained and used without authorisation in connection with development of Anthropic’s Claude systems. Reporting on the complaint describes allegations involving scraping, downloading and the acquisition of material from allegedly unauthorised sources.
The legal significance lies in the separation of different stages of the AI-development process. The litigation is capable of requiring the court to distinguish between the legality of obtaining a copy of a protected work, retaining it, processing it during model development and any later reproduction of protected material by the model.
The underlying question. Whether particular methods of obtaining and using copyrighted material for AI development infringe rights independently of the separate question whether model training itself may qualify as fair use.
What this development does not establish. The filing is an allegation, not a finding. No court has held that Anthropic unlawfully acquired the publishers’ works or infringed copyright.
What happens next. The defendants must respond and the action will enter its initial pleading and case-management stages.
Verification: Federal docket confirmed. Case number, filing date and defendants verified.
Musicians bring identity and publicity-right action against Suno
Status: New canonical matter — ALR number to be assigned
Jurisdiction: United States — federal
Court: US District Court for the District of Massachusetts
Case:Lowery and others v Suno, Inc.
Case number: 1:26-cv-14005
Area: Right of publicity · Identity · Generative music
Stage: Proposed class action filed
Event date: 31 August 2026
Confidence: High for filing; high for core reported claims
David Lowery, Jason Isbell and Eduardo Calle have filed a proposed class action against generative-music company Suno.
The federal docket confirms the plaintiffs, defendant, case number and filing date.
The reported complaint differs from conventional music-copyright cases. The plaintiffs contend that Suno has commercially exploited artists’ names, likenesses and identities in ways that allow users to generate material associated with particular performers.
That distinction matters because a copyright licence to a composition or sound recording does not necessarily determine whether a person’s identity may be commercially appropriated.
The underlying question. Whether the design and commercial operation of a generative-music service can engage state-law rights protecting a performer’s identity even where the dispute is not confined to ownership of a particular musical work.
What this development does not establish. No class has been certified and no court has found that Suno unlawfully used the identity or likeness of any plaintiff.
What happens next. Suno must respond and the plaintiffs will have to establish both the substantive rights asserted and the requirements for any class treatment.
Verification: Federal docket confirmed. The report does not attribute claims beyond those presently supported by the filed matter and authoritative reporting.
Gerencia 360 sues Suno and Bright Data
Status: New canonical matter — ALR number to be assigned
Jurisdiction: United States — federal
Court: US District Court for the Northern District of California
Case:Gerencia 360 Music, Inc. v Suno, Inc., Bright Data Ltd. and Bright Data Inc.
Case number: 5:26-cv-09248
Area: Copyright · Data collection · AI training · Data intermediaries
Stage: Complaint filed
Event date: 31 August 2026
Confidence: High
Gerencia 360 Music filed a separate copyright action against Suno and Bright Data on 31 August.
The docket confirms all three defendants, the filing date and the copyright nature of the action.
The addition of Bright Data makes the case particularly relevant to the developing legal treatment of AI data supply chains. It potentially places scrutiny not only on the developer operating the generative system but also on a company alleged to have participated in the acquisition or delivery of source material.
The underlying question. How legal responsibility should be distributed between a company obtaining or facilitating access to source data and a downstream AI developer that uses that material.
What this development does not establish. The filing does not establish unlawful scraping, circumvention or copyright infringement by either Suno or Bright Data.
What happens next. Initial pleadings and case management. The current docket lists an initial case-management conference for 1 December 2026.
Verification: Federal docket confirmed, including the 1 December case-management date.
US Government enters consolidated OpenAI copyright litigation
Status: New event on existing OpenAI copyright litigation
Jurisdiction: United States — federal
Court: US District Court for the Southern District of New York
Matter:In re OpenAI, Inc. Copyright Infringement Litigation, including The New York Times litigation
Area: Copyright · Fair use · Model training · Government intervention
Stage: Statement of Interest filed
Event date: 1 September 2026Confidence: High
The United States Government has filed a Statement of Interest in the consolidated copyright litigation concerning OpenAI.
The Government argues that training large language models using copyrighted material can, depending on the application of the statutory factors, constitute fair use.
The intervention matters because it places the federal Government’s interpretation of copyright directly before the court in litigation that includes The New York Times’ claims against OpenAI and Microsoft.
The Government’s submission connects copyright doctrine with technological development, competition, scientific capacity and national-security considerations.
The underlying question. Whether use of copyrighted expressive material during training of a general-purpose AI model satisfies the US fair-use test.
That question is not necessarily identical across all AI copyright cases. The source from which material was acquired, the character of the works, the purpose of the use, model behaviour, market substitution and licensing markets can all affect the analysis.
What this development does not establish. A Statement of Interest is not a judgment and does not determine that OpenAI or Microsoft acted lawfully.
What happens next. The district court will consider the Government’s position together with the parties’ substantive copyright arguments.
Brazil’s Superior Electoral Court adopts deepfake test for 2026 election
Status: New judicial event / canonical electoral-AI matter
Jurisdiction: BrazilCourt: Tribunal Superior Eleitoral
Area: Electoral law · Deepfakes · Synthetic political media · Political advertising
Stage: Judgment and interpretative test adopted
Event date: 1 September 2026
Confidence: High
Brazil’s Superior Electoral Court has defined the criteria for identifying a deepfake for the purposes of the 2026 election.
The case arose from an AI-generated video recreating the image and voice of former President Jair Bolsonaro during the Partido Liberal convention at which Flávio Bolsonaro’s presidential candidacy was formalised.
By five votes to two, the TSE adopted a definition covering synthetic material produced or manipulated through AI or equivalent technology that has sufficient realism or verisimilitude and creates, reproduces or alters the image, voice or manifestation of a living, deceased or fictional person.
The court separately held that the electoral prohibition on deepfakes requires the content to constitute electoral propaganda.
By four votes to three, the TSE rejected the request for a fine against Flávio Bolsonaro in the particular case. The majority relied on the convention context and the absence of an explicit request for votes.
The underlying question. The judgment separates two inquiries: whether content is technically a deepfake, and whether its use falls within electoral propaganda prohibited by the election rules.
What this development does not establish. The TSE did not authorise unrestricted use of political deepfakes.
What happens next. The test will apply to future disputes concerning synthetic political media during the remainder of Brazil’s 2026 election cycle.
Verification: TSE primary publication reviewed.
PROFESSIONAL RESPONSIBILITY & JUDICIAL ADMINISTRATION
D.C. Court of Appeals strikes brief containing fictitious authorities
Status: New canonical court-administration matter
Jurisdiction: District of Columbia, United States
Court: District of Columbia Court of Appeals
Case:Barry Douglas v Deutsche Bank National Trust Company
Case number: 24-CV-1099
Area: Professional competence · AI-assisted research · Hallucinated authorities
Stage: Published order; appellee’s brief struck
Event date: 3 September 2026
Confidence: High
The District of Columbia Court of Appeals struck Deutsche Bank’s appellate brief after discovering authorities that did not exist.
The published order records that four cited authorities were fictitious. Counsel explained during the show-cause process that Google’s generative-AI search functionality had been used and that the authorities were not independently verified before filing.
On 3 September, the court ordered the brief struck.
The development continues a line of judicial decisions treating AI-assisted legal research through existing duties of competence and verification. The fact that an AI system generated or surfaced an authority does not transfer responsibility for presenting it to the court.
The underlying question. Whether use of an AI research tool changes a lawyer’s responsibility for the accuracy of legal authorities filed with a court.
What this development does not establish. The order does not prohibit use of generative AI in legal research and does not determine the underlying merits of the foreclosure appeal.
Verification: Published appellate order reviewed.
Fifth Circuit considers reassignment after AI-affected judicial order
Status: New event on existing litigationJurisdiction: United States — federalCourt: US Court of Appeals for the Fifth CircuitCase:Jackson Federation of Teachers and others v Lynn Fitch and othersCase number: 25-60496Area: Judicial use of AI · Appellate review · Reassignment · Court integrityStage: Oral argumentEvent date: 31 August 2026Confidence: High
The Fifth Circuit heard argument on 31 August in litigation challenging Mississippi restrictions on diversity, equity and inclusion programmes.
The AI issue arose from an earlier district-court order containing factual and citation errors. District Judge Henry Wingate subsequently acknowledged that a law clerk had used Perplexity during preparation of material associated with the ruling.
Mississippi has asked the Fifth Circuit both to overturn the preliminary injunction and to reassign the case to another district judge.
That distinguishes the proceeding from conventional lawyer-sanctions cases. Where an AI-affected document is itself a judicial order, the potential remedies involve appellate supervision and the institutional standards governing judicial proceedings.
The underlying question. What remedy is appropriate where use of generative AI within the judicial process contributes to substantial error in a ruling under appeal.
What this development does not establish. The Fifth Circuit has not yet ordered reassignment.
What happens next. The court will decide both the substantive appeal and the reassignment request.
Verification: Fifth Circuit oral-argument calendar confirms the 31 August hearing; authoritative reporting confirms the reassignment issue and the earlier Perplexity use.
Ireland’s GenAI Practice Directions take effect
Status: New event on previously tracked court-rule matter
Jurisdiction: IrelandCourts: High Court and Court of Appeal
References: HC 142 and CA 16
Area: Court procedure · Generative AI · Professional responsibility
Stage: Practice Directions operative
Event date: 1 September 2026
Confidence: High
Ireland’s new Practice Directions on responsible use of generative artificial intelligence in court documents came into operation on 1 September.
The directions do not prohibit use of generative AI.
Instead, they preserve the central principle that the person responsible for a court document remains responsible for its content. AI use does not reduce existing duties of accuracy, candour or professional responsibility.
The framework covers material including pleadings, submissions, affidavits, witness statements and expert reports.
The underlying question. How courts can permit AI assistance without allowing use of an automated system to dilute the standards applicable to documents placed before the judiciary.
What happens next. The directions are now operational and can be applied when AI-assisted material creates practical or procedural issues before the Irish courts.
Verification: Courts Service primary material confirmed.
REGULATORS, ENFORCEMENT & OVERSIGHT
European Commission designates ChatGPT under the Digital Services Act
Status: New canonical regulatory matter
Jurisdiction: European Union
Authority: European Commission
Area: Digital Services Act · Systemic risk · Generative AI · Platform supervision
Stage: Formal VLOSE designation
Event date: 31 August 2026
Confidence: High
The European Commission formally designated ChatGPT as a Very Large Online Search Engine under the Digital Services Act.
The designation follows ChatGPT declaring at least 45 million average monthly users in the European Union.
As a designated service, ChatGPT becomes subject to the DSA’s additional obligations concerning systemic-risk assessment and mitigation.
Those risks can include illegal content, protection of minors, fundamental rights, physical and mental well-being, electoral processes, public security and risks associated with recommender or algorithmic systems.
What this development does not establish. Designation is not a finding that ChatGPT has breached the DSA.
What happens next. The Commission states that the additional obligations apply four months after notification. The precise January 2027 endpoint should not be stated more narrowly until the formal notification date is reconciled.
Verification: Primary European Commission source located in the underlying verified register.
Financial Stability Board raises frontier AI as a financial-system cyber risk
Status: New canonical matterJurisdiction: International / G20Authority: Financial Stability BoardArea: Financial stability · Frontier AI · Cybersecurity · Operational resilienceStage: Formal policy communicationEvent date: 31 August 2026Confidence: High
FSB Chair Andrew Bailey formally raised frontier-AI capability with G20 Finance Ministers and Central Bank Governors ahead of their 31 August–1 September meetings.
The FSB identifies cyber risk as the most immediate potential financial-stability issue arising from advanced frontier models.
Its concern is not restricted to financial institutions directly deploying AI. The letter considers whether increased model capability could alter the speed, cost and scale of malicious cyber activity affecting financial institutions and critical third-party service providers.
What this development does not establish. The FSB has not adopted a binding frontier-AI standard, prohibited any model or identified a particular developer as a systemic threat.
What happens next. Any later FSB, G20 or national supervisory measures will be recorded as subsequent events.
Verification: Direct FSB document confirmed in the verified weekly register.
China reports large-scale AI-app enforcement action
Status: New enforcement event on existing Qinglang workstreamJurisdiction: People’s Republic of ChinaAuthority: Cyberspace Administration of ChinaArea: AI-generated content · Impersonation · Online safety · Platform enforcementEvent date: 2 September 2026Confidence: High
China’s Cyberspace Administration reported the second phase of its Qinglang campaign targeting disorder in AI applications.
The CAC says more than 5.61 million pieces of unlawful or harmful information were removed, action was taken against more than 49,000 accounts, and more than 2,400 websites or applications were dealt with.
The campaign targets categories including AI-generated false information, impersonation, harmful material involving minors and other non-compliant AI applications.
What this development does not establish. The aggregate figures are campaign statistics, not 49,000 individually adjudicated AI-law cases.
Verification: CAC primary publication confirms the figures. (China Anti-Corruption Agency)
Texas confirms operational TRAIGA consumer complaint mechanism
Status: New event on existing Texas AI statute
Jurisdiction: Texas, United States
Authority: Texas Attorney General
Instrument: Texas Responsible Artificial Intelligence Governance Act
Area: AI governance · Consumer complaints · Enforcement implementation
Stage: Complaint mechanism operational
Date: Confirmed operational by 2 September 2026
Confidence: High
Texas has moved further into implementation of TRAIGA.
The Attorney General now provides a Consumer AI Rights resource and a route through which members of the public can submit complaints concerning conduct potentially falling within the statute.
The mechanism was required under TRAIGA following the Act’s commencement earlier in 2026.
What this development does not establish. The existence of the portal is not itself an enforcement action and does not establish any violation.
What happens next. The first identifiable investigation or enforcement action arising through the statutory framework will be the next substantive event.
LEGISLATION & RULEMAKING
California Legislature passes SB 813 on independent AI verification
Status: New legislative event
Jurisdiction: California, United States
Reference: SB 813
Area: AI safety assessment · Independent verification
Stage: Legislature passed bill; enrolled; gubernatorial action pending
Event date: 30 August 2026
Confidence: High
California’s Legislature completed passage of SB 813, which would establish a state framework relating to independent third-party verification of AI systems.
The bill was approved unanimously in the Senate following Assembly passage and was subsequently enrolled.
The framework concerns organisations carrying out independent assessments of AI systems and potential safety risks.
What this development does not establish. SB 813 had not become California law at the reporting cut-off.
What happens next. Gubernatorial action.
Verification: California legislative material confirms passage on 30 August and enrolment on 1 September. (Senator Jerry McNerney)
California Legislature passes No Robo Bosses Act
Status: New legislative event
Jurisdiction: California, United States
Reference: SB 947
Area: Employment · Automated decision systems · Human oversight
Stage: Legislature passed bill; Governor’s consideration pending
Event date: 31 August 2026
Confidence: High
California lawmakers approved SB 947, the No Robo Bosses Act of 2026.
The bill would prohibit employers from relying solely on automated decision-making systems when firing or disciplining workers and would require human oversight and verification where an automated system contributes to those decisions.
The bill also contains restrictions concerning certain predictive uses of workers’ personal information.
What this development does not establish. Legislative passage is not enactment. The measure remained subject to gubernatorial action at the reporting cut-off.
Verification: California Senate primary publication confirms final passage. (Senator Jerry McNerney)
Australia opens second-stage privacy reform consultation addressing AI-era risks
Status: New canonical public-law reform matter
Jurisdiction: Australia — federal
Authority: Attorney-General’s Department
Area: Privacy · AI · Smart glasses · Connected systems · Consent
Stage: Draft legislation and consultation
Event date: 31 August 2026
Confidence: High
The Australian Government opened consultation on further reforms to the Privacy Act.
Government material expressly identifies emerging technologies including artificial intelligence, smart glasses and AI-enabled connected vehicles among the conditions motivating the reforms.
Proposals include stronger tests governing collection and use of personal information, consent requirements and additional rights relating to digital platforms.
What this development does not establish. The proposals are draft reforms and are not yet enacted law.
What happens next. Consultation submissions close on 18 September 2026.
Verification: Primary Australian Government documentation confirmed in the verified register.
US lawmakers announce proposed Ban Artificial Superintelligence Act
Status: New proposed legislative matter
Jurisdiction: United States — federal
Sponsors: Senator Bernie Sanders and Representative Greg Casar
Area: Advanced AI · Artificial superintelligence · Federal AI regulation
Stage: Proposed legislation announced; formal congressional introduction not yet verified
Event date: 3 September 2026
Confidence: High
Senator Bernie Sanders and Representative Greg Casar announced forthcoming legislation that they say would prohibit development and deployment of defined artificial superintelligence and temporarily pause specified advanced-AI development pending federal safety rules.
The sponsors also propose a federal AI regulatory body and international measures intended to prevent artificial-superintelligence development.
The procedural status is important.
At the reporting cut-off, the sponsors’ own material described the legislation as forthcoming. No introduced bill number had been verified.
What this development does not establish. There is no enacted prohibition and the proposal should not yet be described as an introduced Act of Congress.
What happens next. Formal introduction and publication of bill text.
South Korea prepares subordinate rules under AI Data Centre legislation
Status: New legislative implementation event
Jurisdiction: South Korea
Authority: Ministry of Science and ICT
Area: AI infrastructure · Data centres · Subordinate legislation
Stage: Draft Enforcement Decree; public hearing announced
Event date: 4 September 2026
Confidence: High
South Korea announced a public hearing on draft subordinate legislation under the AI Data Centre Industry Promotion Special Act.
The proposed Enforcement Decree is not yet operative law.
Registration for the hearing closes on 7 September and the hearing is scheduled for 9 September.
The underlying question. AI governance is moving beyond software and model obligations into the physical compute infrastructure required to deploy AI at scale.
What happens next. Public hearing followed by the formal legislative-notice process.
GOVERNMENT & PUBLIC AUTHORITIES
UK opens first Sovereign AI R&D procurement competitions
Status: New canonical procurement matter
Jurisdiction: United Kingdom
Authorities: HM Treasury · Cabinet Office · Sovereign AI · participating departments
Area: Sovereign AI · Public procurement · Government deployment
Stage: First competition round launched
Event date: 31 August 2026
Confidence: High
The UK Government launched the first competitions under its £100 million Sovereign AI R&D Procurement Scheme.
The programme targets demonstrator-stage British AI technologies capable of being tested in real-world public-service and national-capability environments.
Initial areas include healthcare, cyber and national-security applications.
What this development does not establish. Opening a procurement competition is not an award to any company.
What happens next. Individual competitions will proceed to assessment and procurement decisions.
Verification: GOV.UK primary material established in the verified register.
Egypt and Cisco sign AI strategy cooperation letter
Status: New public-authority cooperation matter
Jurisdiction: Egypt
Authority: Ministry of Communications and Information Technology
Other party: Cisco Systems
Area: National AI strategy · Government adoption · Digital transformation
Stage: Letter of intent signed
Event date: 1 September 2026
Confidence: High
Egypt’s Ministry of Communications and Information Technology signed a letter of intent with Cisco Systems aimed at supporting implementation of Egypt’s National AI Strategy and other priority digital initiatives.
The Government says cooperation will explore expansion of AI solutions across government and priority sectors to improve service delivery, operational efficiency and decision-support processes.
What this development does not establish. The letter of intent is not evidence that a particular Cisco AI system has been procured or deployed.
Verification: Egyptian State Information Service primary publication confirmed. (SIS)
G20 Innovation Ministers agree AI and emerging-technology principles
Status: New international policy matter
Jurisdiction: G20
Authority: G20 Innovation Ministers
Area: AI policy · Intellectual property · Standards · Workforce
Event date: 2 September 2026
Confidence: High
G20 Innovation Ministers concluded their meeting in North Carolina with a consensus statement addressing emerging technologies.
The six agreed pillars expressly include intellectual-property policy for artificial intelligence and “AI for standards and standards for AI”, alongside technical workforce development and broader pro-innovation policy.
Ministers also agreed the Carolina Principles for Emerging Technologies.
What this development does not establish. These are intergovernmental policy principles, not directly enforceable domestic legislation.
Verification: White House primary publication. (The White House)
UAE deploys 32 specialised Cabinet AI advisers
Status: New public-sector AI deployment matter
Jurisdiction: United Arab Emirates
Authority: UAE Cabinet
Area: Agentic AI · Government decision support · Legislative analysis
Stage: Launch and use approved
Event date: 2 September 2026
Confidence: High
The UAE Cabinet approved the launch and use of a Cabinet AI Advisor system consisting of 32 specialised AI advisers.
Government material says the advisers analyse policies, legislation and initiatives, evaluate financial, economic, social and environmental impacts, compare proposals with international practice and provide recommendations.
The Cabinet also approved a governed process for issuing Cabinet decisions supported by agentic-AI technologies.
The distinction between support and authority remains important. The official account describes AI as supporting government decision-making within a governed framework rather than independently exercising Cabinet powers.
Verification: UAE Government primary publication reviewed. (Government of Dubai Media Office)
UAE expands national AI curriculum
Status: New event on national education-AI programme
Jurisdiction: United Arab Emirates
Authority: UAE Cabinet
Area: Education · AI literacy · Public policy
Event date: 2 September 2026
Confidence: High
At the same Cabinet meeting, the UAE approved wider implementation of its national AI curriculum across public and private schools following earlier pilot work.
The curriculum addresses AI foundations, data, algorithms, ethics, applications and risk.
More than 22,000 teachers and educators are intended to receive associated AI capability development.
This is national education policy rather than commercial AI regulation.
New York City restricts pupil-facing generative AI for younger students
Status: New administrative policy matter
Jurisdiction: New York City
Authorities: Mayor of New York City · New York City Public Schools
Area: Education · Generative AI · Children
Stage: One-year moratorium announced
Event date: 2 September 2026
Confidence: High
New York City announced a one-year moratorium on student-facing generative AI for pupils from 2-K through eighth grade during the 2026–27 school year.
The policy affects nearly 600,000 pupils.
High-school students will instead receive AI critical-thinking instruction and participate in more limited supervised AI pilots.
What this development does not establish. This is not New York State legislation and is not a blanket prohibition on all educator uses of AI.
Canada launches Responsible Data Centre Development Principles
Status: New national AI-infrastructure policy matter
Jurisdiction: Canada
Authority: Innovation, Science and Economic Development Canada
Area: AI infrastructure · Data centres · Energy · Water · Sovereign compute
Stage: Voluntary national principles launched
Event date: 3 September 2026
Confidence: High
Canada launched national principles covering responsible development of data centres.
Government material links domestic data-centre capacity directly to AI compute requirements, sovereign AI infrastructure and wider national resilience.
The principles address local benefits, energy costs, environmental impacts, water use, transparency and strategic value.
What this development does not establish. The principles do not replace existing federal, provincial, territorial, municipal or Indigenous regulatory processes and are not a new federal licensing regime.
Salta confirms AI agents operating in provincial administration
Status: New public-sector deployment matter
Jurisdiction: Argentina — Province of Salta
Authority: Government of Salta
Area: AI agents · Public administration · Government automation
Stage: Deployment publicly confirmed
Event date: 3 September 2026
Confidence: High as to the Government’s claim; technical details incomplete
The Government of Salta announced that AI agents have been developed and implemented within provincial public administration.
A public presentation is scheduled for 7 September.
The precise systems, providers, data sources, administrative functions and any authority to affect legally consequential decisions have not yet been disclosed.
What this development does not establish. It should not yet be said that the AI agents independently make binding administrative decisions.
Vietnam’s replacement National AI Strategy takes effect
Status: New national AI-policy matter
Jurisdiction: Vietnam
Authority: Government of Vietnam
Reference: Decision No. 1671/QD-TTg
Area: National AI strategy · Public administration · Infrastructure · Skills
Event date: 28 August 2026
Confidence: High
Vietnam adopted a replacement National Strategy on Artificial Intelligence to 2030, with a vision to 2045.
The new strategy replaces the country’s 2021 framework and broadens the emphasis from individual AI research projects and applications toward wider national transformation involving public administration, infrastructure, adoption, governance and skills.
Government material includes objectives concerning routine AI use by public officials and greater use of data and AI in administrative decision support.
What this development does not establish. Strategic targets do not automatically create private-sector offences or enforcement duties.
South Korea selects providers for nationwide “AI for All” services
Status: New event in South Korean public-AI programme
Jurisdiction: South Korea
Authority: Ministry of Science and ICT
Area: Public AI · Government-supported deployment
Event date: 28 August 2026
Confidence: High following reconciliation
Consortia led by SK Telecom, Kakao and KT were selected for South Korea’s nationwide public-facing AI programme.
The programme contemplates general-purpose conversational services and public AI agents, with Government infrastructure support.
The selection is distinct from South Korea’s separate internal-government On-AI programme.
What this development does not establish. Selection is not regulatory certification of any model.
What happens next. Agreements, beta operation and a targeted nationwide launch before year-end.
Philippine National Police prepares AI-enabled body-camera deployment
Status: New public-sector deployment matter
Jurisdiction: Philippines
Authority: Philippine National Police
Area: Policing · AI-enabled sensing · Evidence
Event date: 28 August 2026
Confidence: High for programme existence; technical detail incomplete
The Philippine National Police announced preparations to deploy AI-enabled body-worn cameras.
The systems are intended to support documentation, evidence preservation and accountability during police operations.
The verified material does not establish the precise AI functions.
What this development does not establish. Facial recognition, predictive policing, behavioural scoring or autonomous enforcement must not be inferred merely from the phrase “AI-enabled”.
Hong Kong school privacy-AI sandbox reaches briefing stage
Status: New event on existing sandbox matter
Jurisdiction: Hong Kong
Authorities: Privacy Commissioner for Personal Data · Digital Policy Office
Area: Education · Privacy · Generative AI · Regulatory sandbox
Event date: 28 August 2026
Confidence: High
The scheduled briefing for Hong Kong’s Safeguarding Personal Data AI Sandbox was held on 28 August.
The six-month programme is aimed at publicly funded primary and secondary schools and addresses the use of AI-enabled education technology within Hong Kong’s personal-data framework.
The briefing covered eligibility, application criteria, privacy requirements, security and generative-AI technical guidance.
Applications remain open until 30 October 2026.
European Commission Edge Observatory opens Edge AI workshop process
Status: New public-policy event
Jurisdiction: European Union
Authority: European Commission / Edge Observatory
Area: Edge AI · Distributed infrastructure · Security · Sustainability
Event date: 28 August 2026
Confidence: High
The European Commission’s Edge Observatory announced a stakeholder workshop on Europe’s edge-AI ecosystem.
The programme will consider current deployment, future development and requirements for trusted and scalable distributed AI infrastructure.
The workshop is scheduled for 30 September 2026.
This is a policy and evidence-gathering process, not AI Act enforcement.
Vietnam launches dedicated AI category in Make in Viet Nam awards
Status: New government programme event
Jurisdiction: VietnamAuthority: Ministry of Science and Technology
Area: AI products · Innovation policy · Regulatory sandboxes
Event date: 28 August 2026
Confidence: High
Vietnam launched the 2026 Make in Viet Nam Digital Product Awards with a dedicated Outstanding Artificial Intelligence Product category.
Government-linked material indicates that emerging products operating in areas with incomplete legal frameworks may also be prioritised for relevant policy sandbox mechanisms.
Applications close on 15 October 2026.
An award or participation in the programme does not itself constitute regulatory approval.
BACKFILLED & LATE-IDENTIFIED DEVELOPMENTS
The following matters were identified during this week’s monitoring but the substantive event occurred immediately before, or partly outside, the core reporting window. They are preserved because the Law Report database is intended to maintain an audit trail rather than silently lose late-discovered developments.
FTC finalises Active Listening AI marketing orders
On 27 August, the US Federal Trade Commission finalised consent orders against CMG Media Corporation, MindSift and 1010 Digital Works over representations concerning the purported AI-powered “Active Listening” advertising service.
The FTC’s case was directed at alleged representations about capabilities, voice data, consent and targeting rather than a finding that the underlying technology actually performed as advertised.
The final orders move the matters from proposed settlement into compliance and redress.
The total agreed monetary payments are $930,000.
This is a new event on existing FTC matters first initiated in May 2026.
Five Country Ministerial — AI and national security
Security and home-affairs ministers from Australia, Canada, New Zealand, the United Kingdom and United States met on 25–26 August, with the communiqué publicly released on 28 August.
The governments committed to deeper cooperation around frontier AI, cybersecurity, AI-enabled terrorism, scams and hostile-state threats.
The communiqué is an intergovernmental commitment rather than a source of new domestic statutory powers.
Philippines — CICC / Meta deepfake cooperation
A 27 August meeting between Philippine government representatives and Meta was followed by a 28 August Cybercrime Investigation and Coordinating Center statement describing commitments around harmful AI-generated deepfakes and lawful information requests.
The event is cooperation between government and platform, not an enforcement finding against Meta.
Nigeria Defence Headquarters responds to AI impersonation
Nigeria’s Defence Headquarters issued a statement on 27 August rejecting as fabricated an AI-generated video purporting to show the Chief of Defence Staff making partisan political statements.
The statement confirms the Government’s position that the material is synthetic. It does not identify its creator or establish criminal liability.
CNIL updates Genmod AI-model genealogy demonstrator
On 26 August, France’s CNIL released a substantial update to Genmod, its demonstrator for tracing relationships between open AI models.
The tool is designed in part to help explore how personal-data memorisation may propagate across related models.
This is a technical and supervisory research resource, not an enforcement finding against a model provider.
PARTIALLY VERIFIED / REQUIRING FURTHER DOCUMENTATION
Spain — Data Centre Royal Decree consultation deadline
Spain’s consultation on a proposed Royal Decree dealing with data-centre sustainability, resilience and digital sovereignty is verified, and its AI nexus is clear.
The Government connects data-centre development to national AI strategy and future compute requirements.
The procedural deadline is inconsistent across official Ministry material: the detailed consultation page gives 4 September while the central participation index gives 10 September.
Until a formal extension or correction is located, neither date should be stated as definitively controlling.
The proposal remains draft law.
United Kingdom — digital replicas and personality rights
The UK Government’s wider Copyright and AI workstream continues to include consideration of possible protections relating to digital replicas and personality interests.
Fresh reporting indicates continued work, but no new formal consultation commencing during this reporting period was independently verified.
It remains a continuing workstream rather than a new consultation event.
Mexico — national-security screening of AI-related foreign investment
Reporting indicates that the Mexican Executive transmitted proposals to reform foreign-investment law and strengthen national-security screening involving critical technologies including artificial intelligence.
The proposal is materially supported by multiple Mexican reports, but the primary legislative instrument had not been obtained in the week’s verification process.
It remains Partially Verified and must not be described as enacted Mexican law.
Kenya — ODPC AI chatbot
Kenya’s data-protection authority has an AI-enabled service accessible through its regulatory environment, but the verified record does not yet establish the precise formal launch date, provider, model or full regulatory function.
The matter remains Partially Verified pending first-party technical and launch documentation.
Japan — proposed defence AI capability
Authoritative reporting indicates that Japan’s FY2027 defence budget request includes an AI-supported command or information-integration capability.
The underlying primary Ministry budget document containing the precise AI provision was not retrieved in the verified workflow.
The proposal should therefore be reported only as a partially verified budget request, not as an approved or deployed defence AI system.
DEVELOPMENTS TO WATCH
7 September — Argentina: Salta Government public presentation of its AI agents.
7 September — South Korea: registration closes for the AI Data Centre subordinate-legislation hearing.
9 September — South Korea: public hearing on the draft Enforcement Decree under the AI Data Centre Industry Promotion Special Act.
10 September — Spain: current central Ministry index date for the Data Centre Royal Decree consultation, subject to reconciliation.
10, 17 and 18 September — European Union: further work in the Education Committee on AI skills and competences.
18 September — Australia: submissions close on the second-stage Privacy Act reform consultation.
23 September — Scotland: sentencing in Procurator Fiscal v Scott Sheldon, involving the use of an AI nudification system.
25 September — United States: arraignment currently listed in United States v Ryan Sherman, involving alleged AI transformation of photographs into unlawful synthetic imagery.
30 September — European Union: Edge Observatory Edge AI workshop.
15 October — Vietnam: applications close for the Make in Viet Nam Digital Product Awards.
30 October — Hong Kong: applications close for the Safeguarding Personal Data AI Sandbox.
California: gubernatorial action on SB 813 and SB 947.
Texas: first identifiable TRAIGA complaint-led investigation or enforcement action.
US Congress: formal introduction, bill number and text of the proposed Ban Artificial Superintelligence Act.
Sony Music Publishing v Anthropic: defendants’ response and early case-management activity.
Lowery v Suno: Suno’s response to the proposed publicity/identity class action.
Gerencia 360 v Suno and Bright Data: initial pleadings; case-management conference currently listed for 1 December.
Consolidated OpenAI copyright litigation: judicial consideration of the US Government’s fair-use position.
Jackson Federation of Teachers v Fitch: Fifth Circuit ruling on the substantive appeal and reassignment request.
Brazil: application of the TSE’s new deepfake test during the 2026 election campaign.
Ireland: practical application of HC 142 and CA 16 in AI-assisted court filings.
METHODOLOGY & VERIFICATION NOTE
The AI Law Report is an independent editorial publication from AI Policies UK bringing together judicial, regulatory, legislative and public-authority developments involving artificial intelligence.
Issue 005 covers the period from Friday 28 August to Friday 4 September 2026.
The ordinary Courts & Litigation discovery stream failed during part of this reporting period. That failure has not been treated as evidence that the courts were inactive. The missing period was separately researched and reconstructed before publication.
For the recovered court matters, case identities, dates and procedural events were checked against available court dockets, published judicial material, official court calendars, official electoral-court material and authoritative legal reporting. A complaint is described as a complaint; an allegation remains an allegation; an appellate hearing is not presented as a judgment; and a Government submission is not treated as a court ruling.
The regulator and public-authority sections were reconstructed from the verified and reconciled event register compiled throughout the week. The underlying monitoring record separately records primary-source status, evidential gaps, duplicates, correction candidates and relationships between existing matters and new events.
Where a primary document could not be obtained or official sources conflicted, the matter has been retained as Partially Verified rather than silently promoted to established fact.
Where a development occurred outside the reporting window but was discovered during the week’s backfill process, it is labelled accordingly.
The report does not assign new permanent ALR numbers until the new matters have been reconciled against the canonical case register. Once assigned, an ALR number remains attached to the underlying matter. Later hearings, filings, appeals, orders or enforcement events should be appended to that same matter record rather than receiving a new canonical identifier.
The AI Law Report reports legal and regulatory developments for information and educational purposes. It does not provide legal advice or predict legal outcomes.
Next edition: Friday 11 September 2026.