Issue 5
Viewpoint
regulation · European Union
EU designates ChatGPT a search engine under the DSA; Reddit and Roblox classed as very large platforms
On 31 August, the European Commission designated ChatGPT a search engine under the Digital Services Act (DSA), a first for an AI chatbot. Reddit and Roblox were placed under the strictest rules as very large online platforms.
Who is affected and how
- compliance specialist
- Services built on ChatGPT can expect risk-assessment, transparency and audit requirements, with a four-month preparation period according to the Hungarian press.
- corporate lawyer
- Clients that integrate ChatGPT need to clarify who is responsible for obligations under the DSA.
- citizen
- Users may gain broader complaint and transparency rights on large platforms.
lawsuit · United States
Sony and Warner sue Anthropic over songs allegedly harvested illegally to train Claude
Sony and Warner filed a copyright lawsuit against Anthropic at the end of August, alleging that songs were illegally harvested to develop Claude. The complaint covers tens of thousands of songs, and…
corporate lawyerContracts with AI providers should be reviewed for warranties on the origin of training data and for indemnification.lawsuit · United StatesUS government backs OpenAI and Microsoft in New York Times copyright case
In early September, the US Department of Justice asked the court to rule in favour of OpenAI and Microsoft in The New York Times's copyright case. The move bears on how training language models on…
lawyerFor clients building a fair use argument, the government's position should be worked into the line of reasoning.corporate lawyerFor AI-developer clients, this signal should be considered in assessing training-data risk, but it does not replace a court ruling.litigation · globalNew copyright suits target Suno: Isbell, SOCAN and a Mexican catalogue
In early September, Jason Isbell and other plaintiffs filed a federal lawsuit against Suno over exploitation of performers' identities, while Canada's SOCAN sued over generated outputs. According to…
lawyerIn asserting claims, training, output and use of identity can each be raised as a separate legal basis.law · United StatesCalifornia legislature passes bill regulating lawyers' use of AI
The California legislature passed a bill regulating lawyers' use of artificial intelligence in early September, and it has gone to Governor Newsom's desk. The bill targets flawed filings generated by…
lawyerAI-generated citations in filings must be checked, because the lawyer is responsible for flawed material.corporate lawyerA law firm's AI policy must be aligned with the California rules.compliance specialistThe firm's internal control and training procedures must be documented.Hungary
regulationHungary moves AI market surveillance to the minister for science and technologycompliance specialistOfficial requests and notifications must be addressed to the new competent body.corporate lawyerClient guidance on national AI Act enforcement should be updated to reflect the change of competence.lawyerIn administrative proceedings, competence and the route of appeal should be checked.complianceHungarian media lawyer: not all AI-made content must be labelled, but errors risk billion-forint finescompliance specialistPublishing and marketing workflows need to identify which AI content falls under the labelling obligation.corporate lawyerContent-creation contracts should set out the liability and labelling rules for AI use.data protectionHungarian press analyses: workplace AI monitoring raises data protection questions, workers can seek human review and complain to the data protection authoritycorporate lawyerBefore introducing AI monitoring at work, a data protection impact assessment must be carried out and employees informed.compliance specialistA procedure must be set up so that the right to human review can be exercised.citizenAn employee can request human review and file a complaint with the NAIH.professionPetition for Hungarian authors calls for labelling AI-written text and debate turns to the right to say nolawyerIn advisory work, the TDM exception and the question of reserved rights must be handled separately.
More key stories of the week
ruling · United States
Wisconsin federal judge rules AI-generated child abuse material can be protected by First Amendment
A federal judge in Wisconsin ruled that AI-generated material depicting child abuse can, in certain cases, enjoy First Amendment protection. In a related case, the Seventh Circuit did not extend the…
lawyerIn criminal cases, the free-expression defence is arising in a new way for AI-generated content.regulation · European UnionAI Omnibus shifts AI Act deadlines; critics say core problems remain
The digital omnibus amends the application deadlines of the AI Act, with the deadline for medical devices moving to 2028. Analysts say the package has not resolved the AI Act's fundamental problems.
compliance specialistCompliance timelines must be aligned with the new deadlines, and sector-specific deviations tracked separately.corporate lawyerThe omnibus introduces no separate penalty, so sanction risk should be estimated under the original provisions of the AI Act.case · United StatesChatGPT conversations can become evidence, without legal privilege protection
US federal courts have so far not recognised attorney-client privilege for chatbot conversations, so they can end up in court. Hungarian media cite the case of a teenager whose ChatGPT conversation…
lawyerClients must be told that facts disclosed to a chatbot are not protected from discovery.citizenPersonal and family matters sent to a chatbot can be used as evidence in litigation.compliance specialistThe staff AI-use policy should prohibit entering client data and confidential information.compliance · European UnionAI Act enforcement begins as the AI Office starts requesting information from companies
According to industry reports, AI Act enforcement began in August 2026, and the AI Office has started requesting information. One report says the Commission is taking action against more than 30…
compliance specialistDocumentation should be kept ready for AI Office requests, as penalties can reach up to 7 percent of global turnover.corporate lawyerGPAI providers and deployers should set out their AI Act documentation obligations in contracts.case · United KingdomUK tribunal strikes off lawyer over fake AI-generated citations, a first in the country
The UK Solicitors Disciplinary Tribunal struck off a registered foreign lawyer for using misleading AI-generated citations and references in his own defence. According to Law360, it is the first such…
lawyerAI-assisted citations must be checked in every filing, as striking off is a real sanction.compliance specialistFirm policies should require a mandatory verification step for AI-assisted filings.case · United States5th Circuit weighs reassigning case after judge used AI to draft error-laden ruling
Mississippi has asked the 5th US Circuit Court of Appeals to reassign a case to another judge, saying the judge wrote an "error-plagued" AI-assisted ruling. The court is treating the request as a…
lawyerMotions to correct AI-flawed rulings and to replace the judge are becoming more realistic.regulationQuebec courts say generative AI cannot replace judges' reasoning
Quebec's courts adopted a joint guideline on judges' use of generative AI in early September. According to the position, AI cannot replace judicial reasoning.
lawyerIn filings, lawyers should expect courts to regulate the use of AI assistance.tool · United StatesLegal AI market: Thomson Reuters launches own model, OpenAI eyes legal push, Newcode funding reaches $20M
Thomson Reuters has unveiled a proprietary legal AI model for document review, and OpenAI is preparing legal software integrations. Newcode's Series A brings its total funding to $20M, and Pablo…
lawyerThe range of document review and research tools is growing, so procurement should include comparative testing.corporate lawyerIn contracts with legal technology vendors, data handling and the verifiability of outputs are key.law · United StatesCalifornia legislature passes several AI bills on safety risks, worker neural data and youth protection
At the end of its session, the California legislature passed several AI-related bills, including one on assessing AI safety risks and one banning AI tracking of workers' neural data. The governor…
compliance specialistCompanies providing services in California need to track which bills become law.corporate lawyerProcessing neural data in the workplace may become prohibited, so HR-technology contracts should be reviewed.regulation · globalUS urges hands-off approach to AI regulation at G20 tech meeting as EU prepares new law
On September 1, at the G20 technology meeting, the United States called for restraint in AI regulation, while the EU prepares another piece of legislation. According to the EU's digital chief, AI…
compliance specialistTo prepare for transatlantic regulatory divergence, it is worth building a flexible compliance framework.corporate lawyerFor cross-border AI providers, contractual compliance clauses need to be aligned with the differing regulatory directions.ruling · United StatesJudge refuses to block Minnesota's AI 'nudification' law; xAI challenge fails
A judge rejected Elon Musk's xAI's request to block Minnesota's law restricting AI "nudification" tools. The law can therefore take effect.
corporate lawyerGenerative image providers must check their compliance with state deepfake bans.compliance specialistA procedure for filtering and blocking user outputs is needed in the Minnesota market.case · United StatesAnthropic settlement: publishers' claims about author payouts are inaccurate, Writer Beware says
Writer Beware says publishers are making incorrect claims about authors' payouts and that the claims process is also error-prone. The deadline for disputes and documentation has been extended from 30…
lawyerWhen representing authors, the contractual basis for the publisher and author shares should be re-examined.In brief
perAnother 30 lawsuits filed against OpenAI over ChatGPT use linked to school shootinglawyerA new category of product liability and negligence claims is opening up over AI chatbot outputs.compliance specialistChatbot providers need to document how they handle risky conversations.rulingIndian Supreme Court sets aside ₹425-crore customs penalty over AI-hallucinated case lawlawyerChecking AI-originated citations can be a new argument in appeals against official decisions.perSuno licenses and litigates in parallel, straining its fair use argumentlawyerThe licensing agreements can be cited against the other side's litigation arguments, which should be factored into litigation strategy.lawUS state AI rules: companion bot laws, Connecticut employer rule, Colorado hearingcompliance specialistRequirements on companion bots, employment and Colorado rules need to be tracked separately at state level.corporate lawyerFor clients operating in several states, differences between state rules need to be mapped.professionUKJT legal statement and English tort law analyses address liability for AI harmslawyerIn liability claims, the duty of care of the organisation using AI needs to be analysed.corporate lawyerAI service contracts should clarify how liability for harm is shared.caseDC court faults Deutsche Bank subsidiary's lawyers over AI hallucinationslawyerCourts are now faulting lawyers for AI hallucinations in cases for major clients too, so every citation must be checked.corporate lawyerOutside firms should be asked for a contractual statement on their AI use and a duty to verify.rulingBombay High Court grants actress Shruti Haasan interim relief against AI deepfakeslawyerIn deepfake cases, an application for immediate judicial protection is an effective tool.lawUS state deepfake porn laws differ, and the Take It Down Act is only a startlawyerWhen representing victims, state and federal tools must be analysed together.compliance specialistPlatforms must adapt their takedown and reporting procedures to differing state rules.rulingChinese court sets layered originality test for copyright protection of AI-generated workslawyerThe originality test can be used as a comparative argument in authorship disputes.perUdio says AI training is fair use and damages should be as low as $200 per worklawyerBecause the court earlier rejected the expansion of the claims, filing a new complaint may be the route for the plaintiff side.perLocal newspapers sue OpenAI and Microsoft over use of paywalled articleslawyerCircumvention of a paywall can be a separate basis for a claim alongside the copyright claim.data protectionCouncil of Europe drafts data protection rules for AI chatbots and agentscompliance specialistFor chatbot and agent services, it is worth following the draft's obligations when designing for data protection.rulingFederal court excludes expert over AI-hallucinated citationslawyerSources in expert opinions must be checked before filing, as exclusion puts the evidence at risk.perMeta AI glasses lawsuit expands to bystanders who say they were recorded without consentcompliance specialistFor wearable camera devices, third-party consent must be built into the data protection impact assessment.lawyerIn mass privacy claims, class actions may also come into play.professionWall Street banks push Big Law to cut fees because of AIlawyerIn fee negotiations, lawyers must be able to justify where value-creating human work remains.corporate lawyerOn the client side, AI efficiency gains can be passed on in the form of fee discounts.rulingCalifornia Supreme Court leaves attorney sanctions for AI hallucinations in placelawyerSanctions for erroneous AI citations stand at the highest level too.rulingJudge strikes down Anthropic blacklisting, finds government retaliated over Pentagon criticismcorporate lawyerThe challengeability of exclusion of AI suppliers by the state, and the proof of retaliation, can serve as a new precedent.lawyerIn defending against procurement exclusions, the constitutional retaliation argument can also be raised.rulingCourts award damages for creating deepfakes even without distributionlawyerIn a deepfake case, creation alone can ground a damages claim.citizenA deepfake victim can seek damages even if the content was not distributed.perLawsuit says Meta used Facebook and Instagram images to build facial recognition and generative AI without consentcompliance specialistFor AI development based on social media images, a legal basis for processing biometric data must be demonstrated.citizenLegal claims can be brought against the use of images posted on social media for training purposes.professionLaw schools launch AI courses and training programs for legal educationperEuropean AI copyright claims: key considerations and music-related lawsuitslawyerIn European AI copyright cases, territorial scope and exceptions differ from US fair use.lawSanders and Casar propose ban on superintelligence; House bill would have NIST track self-improving AIcompliance specialistBecause federal evaluation and reporting obligations for frontier models may be introduced, the proposals are worth following.lawGreece adopts Act 5321/2026 on the supervision of AI systemscompliance specialistCompanies operating in the Greek market need to build the new supervisory framework into their AI Act compliance.caseIndiana courts crack down on AI hallucinations in filingslawyerLawyers who file unverified AI output can face sanctions even in local courts.professionEFF and artist groups file amicus briefs in AI copyright caseslawyerThe amicus briefs can be used as model arguments in the fair use debate.regulationIndian copyright office: AI-generated work can be protected, but AI cannot be the authorlawyerAuthor data must be provided accurately when declaring AI-assisted works.toolJapan's Supreme Court to test AI use in civil trialslawNew York lawmakers urge Hochul to sign AI technology billscompliance specialistProviders of services in New York need to follow the governor's decision because of possible new AI rules.caseSecond Circuit cautions lawyer against AI use in $82 million NYC subway injury caselawyerIn high-value cases, AI use is also part of the standard of care.caseAI citation errors: Canadian tribunal orders C$31,150 in costs; State Farm attorney admits AI hallucinationslawyerFaulty AI-generated citations can lead to cost orders and professional consequences.compliance specialistInsurers and client organisations may expect a demonstrable verification procedure.professionGoogle in talks with Hollywood studios over licences for its AI toolsrulingIllinois appellate court imposes harsh sanctions for AI misuse in workplace discrimination caselawyerWhere filings are AI-assisted, sanctions may be imposed for erroneous citations.perTexas company sues OpenAI, alleging ChatGPT infringes five AI patentscorporate lawyerClients of AI providers should have patent indemnification clauses checked as well, not only copyright ones.lawCalifornia amends AI transparency law weeks after it took effectcompliance specialistExisting transparency obligations need to be aligned with the amended text of SB 1000.regulationEN 18286:2026, first European standard supporting AI Act implementation, publishedcompliance specialistCompliance teams should support the conformity of their systems with documentation based on harmonised standards.caseSelf-represented litigant ordered to file on paper after hiding AI prompts in pleadingslawGermany still has no deepfake offence in force; draft §201b remains a billlawyerIn Germany, deepfake victims can currently rely only on existing civil-law and personality-rights remedies.citizenCriminal-law protection against deepfakes is not yet available in Germany.rulingJudge backs use of AI to sift data in McDonald's wage class actionlawyerIn e-discovery, AI-based filtering can be used with court approval.perUS law firm accuses UK AI software firm of unwanted contract renewal when it tried to leavecorporate lawyerIn AI vendor contracts, performance metrics, exit terms and automatic renewal should be negotiated separately.professionNew Jersey attorneys flag risks of clients using AI for legal worklawyerClients should be informed about the risks of AI-generated legal material, including the confidentiality implications.perUber drivers sue over pay they say is pushed down by algorithmlawyerIn claims against algorithmic pay-setting, the right to transparency and human review may serve as a basis.compliance specialistPlatforms need to document how algorithmic pay-setting works.