Issue 5

Aug 31 – Sep 6, 2026 · 56 stories · 8 of which from Hungary

Viewpoint

FSB chair warns G20 that frontier AI cyber risk threatens financial stability

Andrew Bailey, chair of the Financial Stability Board and governor of the Bank of England, warned G20 finance ministers and central bank governors in a letter in August 2026. The most immediate concern for the financial system is the effect of frontier AI on cyber risk.

Aug 31, 2026 · 101 sources

Who is affected and how

banking professional: Cyber-risk plans and vulnerability management must account for frontier AI models in attackers' hands.
chief financial officer: Risk and liquidity scenarios should include a correction in AI valuations and rising leverage, since Bailey says together they could trigger another crisis.
insurer: Cyber insurance exposure and terms should be reassessed in light of AI-driven attacks.
fintech company: Attack scenarios built on frontier AI should be included in security and business-continuity testing.

regulation · global

Visa, Mastercard and Fiserv join Agentic Payments Alliance as EMVCo publishes draft agentic payments framework

Visa, Mastercard and Fiserv have joined the Agentic Payments Alliance as founding members; the group brings together more than 25 organisations and seeks common rules for payments made by AI agents…

banking professionalComments on the EMVCo draft can be submitted until 30 September, and issuers should already consider their customer and agent identification logic.fintech companyFirms should align with the emerging standards or take part in shaping them, because these will determine how agent payments are accepted.supervision · United Kingdom

FCA warns frontier AI could create bottlenecks in fixing cyber vulnerabilities at UK financial firms

On 2 September 2026, the FCA published lessons on a new webpage about how frontier AI affects the cyber resilience of financial firms. The regulator warns that AI-driven attacks could create…

banking professionalVulnerability patching priorities and capacity must be reviewed, because AI speeds up attacks.fintech companyPatching processes and third-party management should be documented in line with FCA expectations.supervision · European Union

ECB tightens AI requirements for banks, giving them until October to adapt

The European Central Bank is tightening its AI requirements for banks, a Swedish outlet reports. Banks have until October to adapt.

banking professionalEuro-area banks must assess by October whether their use of AI meets the ECB's expectations.case · global

Cyber insurers adapt policy terms to rogue AI agents as AI exclusions come into question

In September 2026, cyber insurers are mainly clarifying how existing terms apply to AI-related losses rather than introducing new exclusions. Meanwhile, the Fed has doubled its AI risk warning, and…

insurerPolicy wording must be clarified for losses caused by AI agents, and the question of AI exclusions must be settled in new contracts.chief financial officerCompanies should clarify in writing with their cyber insurer whether the policy excludes losses caused by an AI agent.

Hungary

regulationMNB recommendation and credit scoring: high-risk AI systems at Hungarian banksbanking professionalThe use of AI in credit-scoring models must be reviewed under the new MNB recommendation and the EU high-risk classification.deploymentRevolut's AI assistant AIR launches in Hungary; data-handling fine print draws debateretail customerBefore using the assistant, it is worth reading which financial data it can access.banking professionalCustomer expectations of AI assistants at Hungarian banks are rising.fintech companyRevolut's move in Hungary puts competitive pressure on domestic digital providers.fraudDeepfake video calls from the "bank" and a new face-verification bypass scam in Hungaryretail customerDo not trust a "bank" on a video call: verification is only possible through a call-back initiated via the bank's official channel.banking professionalFace-based identification alone is not enough; the process needs additional verification steps.caseRomanian taxpayer's faulty AI tax advice raises question of who answers for AI hallucination before tax authorityaccountantAI-generated suggestions must be checked, because before the tax authority the expert is the one held responsible.retail customerAn AI chatbot's tax advice is no substitute for an expert's opinion; the cost of an incorrect tax return falls on the taxpayer.surveyNN Insurance: 46% of Hungarian internet users have asked AI a health questioninsurerIn health insurance customer communication and product information, insurers must reckon with customers informing themselves through AI.retail customerHealth information received from AI is no substitute for information from a doctor or an insurer.surveyCorvinus University and UNIQA launch joint research on the future of AI-based organizational operationsinsurerThe results could provide Hungarian data for insurers' organizational and workforce-planning decisions.chief financial officerThe research could offer a benchmark for Hungarian experience with organizational AI adoption.deploymentYettel customers can now ask their bank's AI about their spendingretail customerAnalysing spending with AI gets simpler, but the answers are worth verifying.banking professionalGrowing customer openness justifies faster rollout of AI-based customer service features.deploymentBank Pekao tests AI that analyses customer feedback to improve its appbanking professionalAI analysis of customer feedback can provide a fast feedback channel for mobile banking product development.

In brief

caseDC appeals court faults lawyers for Deutsche Bank subsidiary over AI hallucinationbanking professionalIn foreclosure and debt-collection matters, rules on AI use and verification need to be set with outside counsel.caseBanks lack the data to price in credit risk from AI-driven job lossesbanking professionalIn retail and mortgage lending, employment risks in AI-exposed sectors are hard to model.chief financial officerCredit portfolio risk scenarios need to include an AI-driven labour market shock as a separate item.regulationMAS and industry publish SAFR framework for runtime safeguards on AI agents in financebanking professionalThe SAFR control set can be a starting point for building runtime protection for agent systems.regulationAI Act Annex III point 5(c) brings insurers' pricing AI under strict requirementsinsurerA risk assessment must be prepared for risk-assessment and pricing AI systems because of their high-risk classification.oversightGermany: BNetzA to lead AI oversight, BaFin to address financial-sector AI risksbanking professionalAt German subsidiaries, it must be clarified which authority examines which use of AI.regulationNAIC AI risk evaluation supplement open for comments until 29 SeptemberinsurerThe supplement, which will shape reporting expectations for AI risk evaluation, can be commented on until the 29 September deadline.toolVisa upgrades its transfer-fraud prevention system with AI and Featurespace technologybanking professionalWhen expanding transfer-fraud prevention, network-level, transaction-based scoring tools should be evaluated.retail customerFor customers, successful blocking before a transfer is made is the most important protection.caseLiberty Mutual sells insurance inside ChatGPT; liability remains unclearinsurerWhen selling through an AI channel, insurers must specify in advance who is liable for incorrect information.retail customerWhen buying insurance through an AI interface, customers should check the terms in the insurer's own documents.surveyAt Jackson Hole, AI scenario shows models anticipating central bank decisions and trading on themchief financial officerAI actors that react faster to central bank communication could raise risks to market liquidity and price movements.fraudFree family codeword app targets AI voice-cloning scams; Indian family business loses Rs 11.8 lakhretail customerOn a call, a pre-agreed codeword can be used to verify a relative asking for money.banking professionalA voice call alone cannot serve as reliable authentication for large transfers.caseJPMorgan executive outlines what AI projects must prove to win internal fundingchief financial officerThe criteria the bank applies to fund AI projects can serve as a model for companies' own business cases.banking professionalAI initiatives require a clear case for value and risk.caseGoldman Sachs partner warns of 'huge danger' if AI replaces bankers' reasoning skillsbanking professionalWhen junior work is handed to AI, firms must ensure that analysts' judgment continues to develop.chief financial officerIn AI-assisted analysis, a human professional must take responsibility for the final reasoning and accountability.investmentAnthropic nears $15bn credit facility as Morgan Stanley and Goldman Sachs close to IPO rolesbanking professionalInvestment banks' roles in the listing and in lending could bring significant fee income.chief financial officerShare offerings by AI companies could reshuffle capital-market capacity and investor attention.investmentAI infrastructure on credit: ByteDance secures $29.6bn loan, Nscale closes $3bnbanking professionalLarge-scale lending for AI infrastructure raises questions of concentration and exposure.chief financial officerThe credit risk of AI borrowers needs to be assessed thoroughly in syndicated financing.regulationBermuda Monetary Authority consults on AI guidance for financial sectorinsurerBermuda's insurance and reinsurance players should assess the guidance's proposed requirements early.toolAgentic AI in audit: AuditFile launches a tool suite, Cherry Hill offers AI-led EQAsaccountantBefore adopting workflow-automation tools, firms need to define how outputs are checked and who is accountable.deploymentHong Kong selects 36 AI use cases for financial-sector GenAI trials, including Ant Bankbanking professionalThe selected cases show which GenAI applications supervisors are examining, including liquidity management.fintech companySandbox programmes offer a testing opportunity for GenAI solutions with supervisory support.deploymentTelangana to monitor taxpayers with AIaccountantWith tax-authority AI monitoring, consistency of client data and accuracy of returns matter even more.deploymentSamil PwC launches Korea's first AI accounting and tax service among accounting firmsaccountantAI services from the large networks may signal price pressure and changing client expectations for smaller firms.toolWhatsApp to screen suspected scam messages with on-device AIretail customerOn-device scam alerts can help spot phishing messages, but they do not replace verification with the bank.banking professionalNew protections on messaging platforms may change where scam attempts can be detected.caseCapitec's AI blocked $43 million in fraud losses in a yearbanking professionalThe return on fraud-prevention AI can be measured in losses prevented, which is useful for internal business cases.chief financial officerA documented amount of prevented loss can support the case for AI investment.deploymentM&T Bank expands enterprise AI use after years of technology overhaulbanking professionalModernising the data and systems foundation is a precondition for AI adoption at banks.toolMambu launches Intelligent Core to connect AI to core banking systemsbanking professionalThe AI integration capabilities of core banking vendors are becoming a new factor in decisions on replacing or extending systems.fintech companyAI layers from core system providers create new integration opportunities and new competition.toolAltruist launches AI financial planning agentbanking professionalAs AI agents are built into advisory workflows, the checking of recommendations and the information given to clients need to be regulated.deploymentAustralian AI firms get new partnership program backed by the big four banksfintech companyThe bank-backed program could provide an entry point to procurement at the big banks.banking professionalThe program gives access to AI vendors for banks' development needs.regulationGerman banking association comments on draft EU Cloud and AI Development Actbanking professionalCADA may affect the terms for procuring cloud and AI infrastructure, so the banking sector's position is worth following.caseIn Taiwan, households borrow and remortgage homes to invest in AI stocksbanking professionalHousehold lending carries growing exposure when the purpose of the loan is AI-stock speculation.oversightASIC removes 19,400 AI-driven scam sites in Australiaretail customerThe reliability of ads and investment offers should be checked in official registers.fraudUnico blocks R$19.6 billion in fraud attempts in first half, records wave of 'homemade deepfakes'banking professionalIdentity verification must allow for deepfakes now being made with home tools.fintech companyLiveness checks need continuous testing against new deepfake techniques.surveyAstraeus: AI adoption linked to higher hiring and advisor productivity in wealth managementchief financial officerIn headcount and productivity planning for advisers, AI adoption should not be treated solely as a cost-cutting measure.investmentFélix raises $200m to fund AI-powered product expansionfintech companyCapital going to AI-based products is intensifying competition in cross-border transfers.banking professionalLarge fintech funding rounds put pressure on banks' transfer pricing and speed.

← Issue 4 · Issue 6 →

Issue 5

FSB chair warns G20 that frontier AI cyber risk threatens financial stability

Who is affected and how

Visa, Mastercard and Fiserv join Agentic Payments Alliance as EMVCo publishes draft agentic payments framework

FCA warns frontier AI could create bottlenecks in fixing cyber vulnerabilities at UK financial firms

ECB tightens AI requirements for banks, giving them until October to adapt

Cyber insurers adapt policy terms to rogue AI agents as AI exclusions come into question

Hungary

More key stories of the week

AI agents get wallets as AWS, MoonPay, Alipay and stablecoin payments advance, with compliance lagging

Wall Street banks press law firms to cut fees to share AI efficiency gains

Anthropic unveils commerce agent blueprint with Visa and Mastercard; Mastercard launches agentic cohort

CFO surveys: data privacy ranks first, AI sixth but rising, and ROI remains in question

98% of claims adjusters' AI reviews on Glassdoor are negative as insurers automate claims

India works on small-value UPI payments by AI agents under user-set rules

PCAOB faces pressure to set an audit standard for corporate AI use

Santander and Mastercard report Europe's first payment made entirely by an AI agent; DBS and Stripe partner

Dutch banking association urges lenders to use AI for transaction monitoring; AML human oversight examined

Big banks and big tech clash over regulation of agentic AI

Metropolitan Police: AI-related fraud reports up 395%, losses reach £9.6 million

Wealthtime: 28% of clients use AI tools to support financial advice

In brief