Issue 3
Viewpoint
law · European Union
Digital Omnibus pushes AI Act high-risk deadlines to 2027–2028
The Digital Omnibus on AI is the first substantive amendment to the AI Act since 2024 and postpones compliance deadlines for high-risk systems. Analyses say it is not a blanket postponement: for medical devices the deadline is August 2028, while other obligations remain in force.
Who is affected and how
- compliance specialist
- Compliance plans must be redrawn obligation by obligation, because only the deadline for high-risk systems has moved.
- corporate lawyer
- References to deadlines and provider commitments in contracts must be reviewed.
- lawyer
- Clients must be told precisely which obligation has been postponed and which has not.
lawsuit · United States
Round Hill files $1 billion copyright lawsuits against Suno and Anthropic
Round Hill, one of the independent music publishers, is seeking more than $1 billion in damages from AI music service Suno and from Anthropic, citing copyright infringement. The lawsuits were…
corporate lawyerA licensing risk assessment should be carried out before using AI music and other generative services.lawyerPreparing new types of large-value claims for music rights holders is becoming realistic.regulation · United StatesUS state and federal AI rules collide: Connecticut, Colorado, FTC
US states are adopting strict AI laws on automated decision systems and generative media, but the FTC says federal law could override Colorado's transparency rules. Connecticut's omnibus AI law, SB5…
compliance specialistA map of state-by-state AI requirements should be kept, especially for automated decisions and employment.corporate lawyerGiven the uncertainty over federal preemption, the compliance programme should be designed flexibly.lawyerClients should be told which state rules apply to them.regulation · United KingdomSRA warns of AI hallucination risks after 42 reports of potential misuse
The Solicitors Regulation Authority, the regulator of solicitors in England and Wales, issued a warning notice reminding firms of their professional obligations after 42 reports of potential…
lawyerEvery citation produced with AI must be verified, because responsibility rests with the solicitor.corporate lawyerLaw firms must set out in an internal policy how AI use is checked and how confidentiality is protected.compliance specialistTools' data handling and controls against fake citations must be audited.compliance · European UnionAI Act enforcement and transparency obligations apply from 2 August 2026
On 2 August 2026 the AI Act entered a new phase: the enforcement powers of the European Commission's AI Office began, and the transparency rules took effect. Analyses say fines can reach €15 million…
compliance specialistLabelling and transparency obligations for AI systems must be assessed and documented immediately.corporate lawyerAn inventory of the AI tools used by the company and a allocation of responsibilities must be drawn up.citizenComplaint routes are now available for reporting breaches of the AI Act.Hungary
complianceHungarian trade outlets explain AI content labelling rules: deepfakes, chatbots, Article 50compliance specialistA labelling process for AI-generated text, images and video must be introduced in domestic communications too.corporate lawyerFor customer-service chatbots, users must be made aware that they are talking to a machine.regulationHungary to get new AI authority; AI fraud victims must turn to different agenciescitizenAI-enabled fraud must be reported to the police as a crime, while another authority is competent for biometric cases.lawyerClients must be directed to the appropriate authority depending on the type of AI fraud.compliance specialistThe split of domestic supervisory powers must be built into the compliance plan.professionHungarian lawyers' debate: it is not true that there are no rules for AIlawyerWhen using generative AI, lawyers must apply existing confidentiality and duty-of-care rules rather than cite a lack of regulation.corporate lawyerThe firm's AI policy should be drawn up according to the considerations in the CCBE guidance.data protectionWhen a data protection impact assessment is mandatory for an AI systemcompliance specialistBefore deploying an AI system, the need for an impact assessment must be determined following the decision sequence.corporate lawyerSupplier contracts must specify who carries out and documents the impact assessment.professionFülöp Henrik analysis examines copyright and generative AI: who owns the outputlawyerWhen taking on client work on the legal status of AI outputs, the uncertainty must be made clear.
More key stories of the week
ruling · United States
Federal court rules judicial immunity covers judge who allegedly relied wholly on AI in order
A federal court ruled that judicial immunity protects a Nevada family court judge even if the judge did in fact leave the drafting of the decision to AI. The plaintiff sued Judge Mari Parladé.
lawyerA damages suit over a judge's use of AI is not a viable route; remedies must be sought through appeal.case · United StatesState Farm defense lawyers admit AI generated fake cases cited in Los Angeles lawsuit
State Farm's defense lawyers admitted that they cited non-existent, AI-generated cases in a Los Angeles lawsuit. The case is an insurance dispute over a house fire.
lawyerAI-generated references must be verified at the source before submission.corporate lawyerOutside law firms should be asked by contract to disclose their AI use and to allow verification.ruling · European UnionMunich court rules for GEMA against Suno in AI music case
A Munich court ruled in favour of collecting society GEMA against AI music service Suno. Legal journals regard the decision as significant for AI music and copyright.
corporate lawyerCompanies offering or using AI music in the European market need to reassess licensing risk.lawyerThe German ruling can serve as a reference in copyright disputes in other EU member states.ruling · United StatesCalifornia court sanctions attorney for delegating AI citation checks to a paralegal
A California court sanctioned an attorney who delegated the checking of AI-generated citations to a paralegal. The commentary says responsibility for AI research rests with the attorney.
lawyerVerification cannot be delegated without responsibility: the attorney must personally answer for the citations filed.corporate lawyerThe internal review process must have a designated person responsible.data protection · European UnionDutch regulator fines Uber €825 million over algorithmic deactivation of driver accounts
A Dutch regulator has fined Uber €825 million for letting an algorithm deactivate drivers' accounts. The news was reported on 21 August 2026.
compliance specialistFor automated decisions, affected people must be given access and the possibility of human review.corporate lawyerPlatforms' algorithmic decision-making should be reviewed from a legal perspective, as the size of the fine is high.citizenIf an account is deactivated by an algorithm, human review can be requested.profession · globalAnalysis counts more than 1,600 court cases with fake AI citations, some sanctions in six figures
An analysis found more than 1,600 court cases involving AI-invented citations, and some sanctions run to six figures. Trade articles say RAG-based tools can also hallucinate, and vendor contracts…
lawyerAI tool output must be checked at the source, because the vendor contract does not take over liability.corporate lawyerLiability and error-handling terms must be negotiated separately when buying AI research tools.tool · globalLegal AI firms move to own models: Harvey launches Tenet, another vendor switches to Kimi K3
Harvey unveiled Tenet, its first in-house legal AI model, rather than relying solely on OpenAI and other developers. According to other outlets, legal AI vendors are building their own models to cut…
corporate lawyerWhen a vendor changes models, check where and by which model client data is processed.compliance specialistData-handling and supplier risk related to third-party models should be reassessed.lawyerLawyers should find out which model sits behind the tool they use, because of confidentiality.tool · United StatesHarvey launches next-generation platform built around memory
Harvey unveiled "Harvey II", a new version of its legal AI platform built around memory, which also adds preference settings and context sharing. The announcement was published on 18 August 2026.
lawyerBefore using memory, it must be clarified what client data stays in the system.corporate lawyerConfidentiality controls on the storage and sharing of client data should be reviewed.ruling · United StatesBartz case becomes the benchmark for AI copyright: fair use and $1.5 billion
Judge Alsup ruled that training AI on legally purchased books is transformative fair use, yet Anthropic still agreed to pay authors $1.5 billion over piracy. Several analyses reviewed the topic…
corporate lawyerThe origin of the data source (lawfully acquired or pirated copy) must be documented, because that is where the risk is decided.lawyerIt is advisable to assess the ruling and the settlement separately when advising clients.compliance · United StatesByteDance signs AI copyright pact with Hollywood trade group
TikTok parent ByteDance has agreed with a Hollywood trade group on copyright limits for AI use. The move follows the dispute with Hollywood.
corporate lawyerFor companies developing video and content generation tools, the content of the commitments is worth treating as a reference.ruling · United StatesConnecticut judge sanctions plaintiff over hidden AI prompt in court filing
A self-represented plaintiff in a Connecticut case placed a hidden prompt in a court filing to influence AI, and the judge discovered it and barred the plaintiff from electronic filing. According to…
lawyerIncoming filings from the opposing side should be checked for hidden text before they are fed into an AI tool.lawsuit · United StatesFederal lawsuits allege xAI's Grok was used to create AI child sexual abuse images
A federal lawsuit alleges that xAI enabled the creation of AI images depicting child sexual abuse, and another woman has joined the suit against Grok. xAI blames a photographer in Arkansas.
lawyerWhen representing victims against AI image-generation providers, platform liability is the central question.compliance specialistImage-generation tools need strong content filtering and a reporting channel.citizenIn the case of AI-generated sexual images, victims can take legal action against the platform.In brief
perIllinois lawsuits target use of voice data to train AIcorporate lawyerAI training that uses voice and biometric data must be grounded in a review of consent.lawyerA new group of Illinois biometric privacy claims is worth watching.citizenA lawsuit can be brought over the use of one's own voice data for training.professionNYC Bar issues ethics opinion on AI recording of conversationslawyerBefore AI recording and summarizing, the other party must be informed and the ethical requirements assessed.corporate lawyerAn internal policy is needed for AI processing of negotiation and interview notes.rulingNinth Circuit: AI agents are browser-like tools, but Amazon's case against Perplexity continuescorporate lawyerA company operating an AI agent must expect that the platform may hold it liable over its web access.lawyerThe access and terms-of-use risks that agent-based AI providers pose for their clients need separate assessment.caseExpert witness used ChatGPT to write report defending 3M in deadly explosion lawsuitlawyerWhen reviewing an opposing expert's report, it is worth establishing how far AI was used and what sources it relied on.corporate lawyerWhen engaging an expert, the contract should set limits on the use of AI.rulingNew Mexico Supreme Court removes lawyer from case over AI errors in brieflawyerA brief with AI errors can lead to removal from the case, which also hurts the client.perFlorida asks court to declare OpenAI and Sam Altman a public nuisancecorporate lawyerState public nuisance suits against AI providers represent a new liability risk.lawyerA claim based on public nuisance is a different approach from the product liability route used so far.regulationSouth Korea opens copyright registration to AI-assisted workslawDeepfake takedown law sets 48-hour deadline and fines of up to $53,088compliance specialistPlatforms must run a 48-hour removal process.citizenThe person depicted can ask the platform to remove the content.lawyerA lawyer representing victims can cite the platform's deadline and the fine as grounds.caseTech companies buy rare books and destroy physical copies for AI trainingcorporate lawyerBook acquisition and digitisation processes should be documented from a legal standpoint as well.toolMore legal AI launches: FoxAI, Litera, DISCO, Filevine, Thomson Reuters–Laurel, Sansancorporate lawyerFor contract and review tools, data handling and liability should be compared before purchase.lawyerThe output of AI used in document review must be checked by human reviewers.toolAderant links iTimekeep to Harvey to turn AI-assisted legal work into billable timelawyerTime recording for AI-assisted work must be reconciled with the fee terms agreed with the client.corporate lawyerThe firm's billing model must settle how time spent working with AI is charged.professionVos: people likely to accept AI judges for small disputeslawyerIn arbitration, it is worth following the soft-law frameworks on transparency and human oversight.toolHarvey picks DeepL for legal translationlawyerIn cross-border matters, translation can be done within the platform, but quality control remains the lawyer's responsibility.corporate lawyerTranslation data flows should be reviewed for data protection and confidentiality.regulationRules on AI political deepfakes differ from state to state in the UScompliance specialistDeepfake labeling and ban rules must be checked state by state for campaign materials.lawyerBecause state rules differ, political clients need a legal pre-screening.perAI and copyright: Thomson Reuters win, ANI v OpenAI, Apple shareholders, Börsenvereincorporate lawyerAI copyright claims can also create executive liability and disclosure risk.regulationLimits of unauthorized-practice rules in regulating mental health AIcorporate lawyerLicensing rules for mental health AI products need to be reviewed state by state.compliance specialistProduct information and limitations for health-related AI features should be checked.regulationIndiana court rule and Maryland standing order set expectations for AI use in filingslawyerA court's AI disclosure and verification requirements should be checked before filing.regulationOpenAI calls for stronger AI laws in Californiacompliance specialistThe direction of California's regulation is worth following, as developers themselves are asking for stricter rules.caseAI slop is swamping the US House office that drafts lawscaseUnverified Jennie clip fuels deepfake fears and legal questions in KoreacitizenA deepfake victim's legal steps should begin with reporting the content to the platform and filing a complaint with the authorities.professionWhen AI designs a drug, who gets the patent credit?lawyerPatent filings should document the human inventor's contribution.corporate lawyerInventor status needs to be clarified when patenting results of AI-assisted research.per"Italian brainrot" dispute raises question of who owns AI-created characterslawyerThe outcome of the case may shape the legal protection of AI characters.professionAI is reshaping entry-level legal jobs for graduates, not eliminating themregulationEurope approves first AI Act standard, but compliance benefit is yet to arrivecompliance specialistUntil then, compliance documentation built on the standard should keep an independent method of demonstrating conformity.corporate lawyerContractual use of presumptions that refer to the standard should be planned only once the legal effect is clarified.caseSpotify moves against non-existent AI artistsprofessionOntario's first AI discipline ruling sanctions the deception, not the chatbotlawyerDisciplinary risk stems from failing to verify and from misleading the court, not from using AI as such.regulationVietnam: fully AI-generated song has no copyright, rightsholders get opt-outprofessionInsurers show growing interest in AI coverage exclusionscorporate lawyerAI exclusion clauses in liability insurance policies should be reviewed before renewal.compliance specialistInsurance coverage for AI risk should be built into the risk management plan.caseAfter a major win, some lawyers back litigants who use AI in courtlawyerWith clients litigating on their own using AI, the value proposition of representation services needs to be rethought.data protectionEuropean data protection regulators set up a ChatGPT task forcecompliance specialistChatbot providers' data processing must be managed with coordinated EU regulatory scrutiny in mind.