Issue 14 · Nov 2–8, 2026Agent Age News magazine · Other editions · Method · HU / ENG

The impact of AI on banks, insurers, pay and finance, curated weekly

Issue 12

Oct 19–25, 2026 · 44 stories · 7 of them from Hungary
Viewpoint

case · United States

US big banks block third-party AI agents from accessing customer accounts

According to Axios, Morgan Stanley, Bank of America, JPMorgan and Wells Fargo detected AI agents trying to act in accounts without customer instructions. The banks are restricting third-party agents; among their concerns is deposits being steered toward higher yields, and Sanders called for a pause on AI.

Oct 19, 2026 · 68 sources

Who is affected and how

retail customer
Do not hand your banking login details to an AI agent, because banks may treat such access as prohibited or unauthorised activity.
banking professional
Recognising agent traffic and tightening customer authentication rules is a direct operational task.
fintech company
Agent-based services must find an authorised API channel instead of the screen- and password-based access that banks prohibit.
regulation · United Kingdom

FCA urges AI providers to warn users about limits of chatbot financial advice

Jessica Rusu, the FCA's chief data and information officer, said at Bloomberg's London series on 19 October that AI providers should warn users about the limits of financial advice from chatbots. The…

retail customerA chatbot's answer is not the same as regulated financial advice, so it should be checked before any major decision.fintech companyFirms offering chatbot-based financial guidance must build in a clear warning about its limits.
deployment · global

Mastercard opens early access to Agent Pay for Machines for AI agents

Mastercard opened early access on 19 October to Agent Pay for Machines, a tool that lets autonomous AI agents buy digital services. It was also opened to developers at Cloudflare Connect, where a…

fintech companyAgent Pay and agent tokens can be integrated for services paid for by agents.banking professionalIssuers must prepare for agent tokens and for authorisation rules covering agent-initiated transactions.chief financial officerMachine purchases need a spending limit drawn from a single budget and an approval policy.
supervision · United Kingdom

FCA sets out good and poor AI practice; firms overestimate model accuracy

The UK financial regulator published good and poor practice for AI use by financial services firms on 19 October. It found that AI governance has lagged behind adoption, that firms overestimate model…

banking professionalClaims about model accuracy must be backed by documented testing.chief financial officerThe governance framework for AI deployments must be settled before go-live, not afterwards.fintech companyFirms must demonstrate that customer interests come first in new AI features.
layoffs · United Kingdom

Barclays to cut 5,000 jobs, about 5% of its workforce, as it steps up AI adoption

Barclays announced on 22 October that it will cut 5,000 jobs, about 5% of its workforce, citing stepped-up AI adoption. The chief executive called AI "the icing on the cake", and according to Reuters…

banking professionalSupport and operations roles are the most exposed, so upskilling and retraining for these roles is urgent.chief financial officerIn cost-cutting programmes, AI-linked savings targets need to be backed by real productivity data.

Hungary

survey48% of Hungarians use AI for financial questions but seek human advice on big decisionsretail customerAI can be used for orientation, but a significant financial decision requires a professional's opinion.banking professionalCustomers increasingly arrive with preliminary information obtained from AI, which has to be handled in client service.fintech companyIn AI-based financial guidance, trust and a human handover point are the competitive advantage.fraudErste Bank customers targeted nearly 21,000 times in six months, with losses of HUF 7 billionretail customerThe customer-approved transfer is the main point of attack, so every unexpected payment request must be verified.banking professionalFraud prevention must also be ready for transactions that are carried out by the victim and look legitimate.caseHungarian banks and payment players on AI-agent shopping: OTP, SimplePay, Woltbanking professionalIn agent-based shopping, the bank's role and its proximity to the customer are a strategic decision.fintech companyPayment providers need to develop an authorisation model for payments initiated by an agent.caseWho pays when AI makes a mistake remains an open question in Hungary; insurers' use of AIchief financial officerThe terms of company liability insurance should be reviewed for damage caused by AI.insurerPolicy terms must clarify coverage for damage arising from AI errors.deploymentK&H's Kate assistant interprets customer requests in conversation and can initiate transfersretail customerIn the conversational banking interface, a transfer can be requested in natural language, but the details must be checked before it is executed.banking professionalConversation-based transactions require rules for authentication and error handling.oversightHungary's tax authority plans AI selection of risky invoices in real time in 2026, accountant saysaccountantThe accuracy and consistency of invoice data come to the fore, because AI screening can flag discrepancies in real time.chief financial officerThe quality of company invoicing data and the handling of risk flags should be sorted out in advance.toolHungarian 25-year-old builds Azta, an AI for legal and tax questionsretail customerAI tools aimed at legal and tax questions are now available, but their answers should be checked.accountantClients increasingly use AI tools to get initial information on tax questions.

More key stories of the week

case · global

Credit markets look for weak links in AI debt as credit spreads hit a 17-month high

According to Bloomberg, fears over AI technology pushed global credit spreads to a 17-month high. Trading in credit default swaps (CDS) rose on AI-related debt, and CoreWeave's bonds came into the…

chief financial officerCredit risk and refinancing should be reassessed in portfolios exposed to AI infrastructure borrowers.banking professionalExposure to AI data-centre financing should be reviewed against concentration limits.insurerCredit risk in the AI sector also shows up in investment portfolios.
tool · global

World launches World ID suite for banks to counter AI-driven fraud and deepfakes

World announced its World ID suite for banks on 19 October and made deepfake verification of Zoom calls generally available. The company reported five development partners but has not yet named a…

banking professionalCustomer identification and video-call authentication can be supplemented with proof of human presence.fintech companyIdentity verification is a new service layer on which a firm's own fraud-prevention solution can be built.
fraud · global

AI voice-cloning and fake-retailer scams target retail customers

US and UK reports say scammers use cloned voices, fake retailers and people posing as police officers or relatives. The UK's National Trading Standards warns of AI-driven fraud; one defense is a…

retail customerAgree a family code word, and when asked for money, call the person back through an independent channel.banking professionalTargeted warnings and delays can be applied to urgent transfers initiated by customers.
deployment · global

Stablecoins compete for AI-agent payments, with USDC in the lead

Analyses say USDC leads among stablecoins in AI-agent payments, and usage could begin as early as 2027. A South Korean letter of intent (between Square Tech and GOAT Network) on agent-based…

fintech companyIn the market for micropayments between agents, stablecoin rails are a competitive alternative to card-based solutions.banking professionalBanks should assess which payment flows could move off bank rails through agents and stablecoins.
tool · global

AI agents for tax and accounting launch: Thomson Reuters, CCH Axcess, Wolters Kluwer

Thomson Reuters has launched AI for tax and audit that shows its reasoning alongside the answer. CCH Axcess has added agentic tax return preparation, and Wolters Kluwer is also introducing an agentic…

accountantAgentic preparation replaces repetitive work, while review and professional responsibility remain with the accountant.chief financial officerTax AI that works with reasoning is more auditable, but its output still has to be reviewed.
deployment · global

Visa, BILL and Billtrust pilot AI agents for B2B invoice payments

Visa is building on AI-based fraud scoring, agent-initiated payments and tokens, and is also opening up to stablecoin rails in B2B payments. Darren Parslow, Visa's global head of commercial…

fintech companyFintech firms that want to take part in agentic payments need to join Visa's agent and token offering as partners.chief financial officerIn B2B payments, stablecoin rails and agent-initiated payouts bring both new opportunities and new risks.
oversight · Asia

Bank of Japan, MAS and China's central bank focus on AI risks and strategy

On October 23, the Bank of Japan warned of the dangers of an AI market shock and AI-driven cyber risks. Singapore's MAS cited financial stability risks from the billions flowing into AI, and China's…

banking professionalScenarios of AI-driven cyberattacks and market shock may be included in risk stress tests.
layoffs · global

Banks and central banks examine AI's impact on employment

ABC asked whether AI could be behind job cuts at NAB and other big Australian banks, while European banks are quietly cutting white-collar staff. Goldman says AI could displace one in 11 workers, but…

banking professionalLabour-market risk is greatest for routine back-office tasks, so it is worth building skills ahead of time.chief financial officerIn workforce planning, AI savings and the economic cycle together shape the pace of layoffs.
fraud · global

Deepfake executive scams: $95 million loss, FBI data, INTERPOL and KeyBank warnings

The articles say Fideuram's deepfake loss of 95 million and the FBI's report of $893 million in AI-assisted fraud show the risk of executive voice cloning. INTERPOL says deepfakes can defeat familiar…

chief financial officerFor large transfers, a voice or video call alone is not sufficient approval; independent call-back verification is mandatory.banking professionalSignals of executive-impersonation fraud against customers must be built into transaction monitoring.
rollout · United Kingdom

Morgan Stanley launches UK savings accounts as US banks fear AI agents could drain cheap deposits

Morgan Stanley announced on Monday that it will enter the UK retail savings market with fintech partner Raisin UK, offering interest of up to 5.1%. US banks are worried that AI agents could draw away…

retail customerCustomers can move money more quickly to accounts offering higher interest.banking professionalAI agents could weaken deposit loyalty, so banks should prepare for rate competition.chief financial officerDeposit-cost and deposit-retention models should also be tested against agent-driven reallocation scenarios.
supervision · United Kingdom

Bank of England considers AI stress test; pension trustees urged to test concentration risk

According to Bloomberg, the Bank of England is considering stress testing the impact of AI on the financial sector. The UK pensions regulator has asked pension trustees to test for AI concentration…

chief financial officerAI concentration should be tested in pension and investment portfolios.
tool · United States

FIS and Anthropic's Claude-based money laundering investigation agent is available to any U.S. bank

On October 19, FIS made an AI agent for money laundering investigations, built on Anthropic's Claude, commercially available to any U.S. bank. The tool reduces the time investigators spend assembling…

banking professionalInvestigators' work shifts toward reviewing the material the agent assembles.fintech companyAML agents from large vendors put pricing and competitive pressure on smaller RegTech players.

In brief

deploymentEuropean financial firms move to open AI models run on their own infrastructurebanking professionalOpen models run on in-house infrastructure give better data control, but the bank is responsible for operating them.fintech companyOpen models can reduce dependence on closed-model providers.surveyFitch sees moderate AI disruption risk for wealth managers and retail banks; BCG says finance is most exposedbanking professionalIn retail banking, AI-based advice creates competitive pressure.chief financial officerAssessing AI exposure is part of strategic planning.oversightUS regulators expect banks to account for AI use; SEC to focus on AI in examsbanking professionalDocumentation on AI system vendors and security should be put in order.regulationHKMA launches "ART" framework for responsible AI adoption in banking; banks test deepfake detectionbanking professionalThe framework for responsible AI adoption can serve as a reference for internal governance.caseAI exclusions spread in insurance; Harbor raises $10 million to cover the gapinsurerExcluding AI risk or making it a standalone product is a strategic decision in policy terms.chief financial officerCompanies should check whether their business insurance excludes losses arising from AI errors.investmentUS life insurers' exposure to the AI sector rose 24% from January to JuneinsurerThe growing weight of the AI sector must be treated separately in investment risk management and capital requirement calculations.toolAI adoption outpaces governance, raising audit risks as AICPA and AIUC launch AI assurance offeringaccountantUsing AI agents in audit and advisory work requires professional judgment and human oversight.chief financial officerThe auditability of AI-based processes must be ensured before they are deployed.deploymentJPMorgan's private bank uses AI to make advisers more efficient; Goldman offers small businesses an AI packagebanking professionalAdvisory work is increasingly AI-assisted, and compliance with security rules is a key issue.chief financial officerBanks also see a new business line in AI-adoption services for small-business clients.deploymentAI cuts tax inspection time in Azerbaijan by 7–9 weeks; Ghana's tax authority reports GH¢1bn gainaccountantAI-based inspections by tax authorities point to faster data requests and more consistent discrepancy detection.chief financial officerAccurate and timely data reporting matters even more in a shorter inspection cycle.fraudDeepfake scammers use faces and voices of central banks and officials in Romania, Malaysia and Indiaretail customerInvestment and benefit offers made in the name of authorities or leaders must be verified through official channels.deploymentUAE tax authority launches AI platform to cut tourist VAT refunds to one dayfintech companyThe digital AI-based refund and payment process creates an opportunity for new services.chief financial officerFaster cross-border VAT refunds affect cash flow and merchants' processes.caseBanks rein in AI costs and turn to cheaper Chinese modelschief financial officerControlling AI costs and choosing models both carry supplier and geopolitical risk.deploymentHuntington Bank launches AI business-startup platform; APG Federal Credit Union goes live with Scienaptic AIbanking professionalAI-supported credit-assessment and small-business onboarding processes are available to smaller institutions too.fintech companyBanks and credit unions are looking for AI decision-support vendors.deploymentBank of Ireland accelerates AI in fraud detection, with 1 billion checks and €9.7m savedbanking professionalAI-based transaction checks can deliver measurable savings at high volume.deploymentBank AI projects: OpenAI on COBOL migration, Danske Bank API agents, BNK group-wide platformbanking professionalAI-assisted migration of legacy core banking code and agent-API integration bring both operational risk and opportunity.fintech companyAgents connecting to banking APIs open up a new integration market.surveyEOS survey: 46% of European consumers would discuss sensitive banking matters with AIbanking professionalA significant share of customers are willing to raise sensitive topics through an AI channel, so data protection and handover to a human agent matter.fintech companyIn sensitive financial situations (for example debt management), an AI channel may be in demand, but trust is critical.toolVeryfi launches AI suite to screen checks, invoices and receipts for document fraudaccountantThe authenticity of submitted invoices and receipts can be checked automatically.banking professionalReal-time API-based screening against check and document fraud can be built in.fraudBanks call for stronger intelligence sharing to fight AI fraudbanking professionalFraud indicators must be shared between institutions, which requires data-protection frameworks.regulationUK tax authority moves to tackle rogue tax advice from artificial intelligenceretail customerAI tax advice should be checked against official sources.accountantClients' incorrect tax advice taken from AI must be corrected.fraudCyberCartel reportedly used Claude and Grok AI agents to automate attacks on Latin American banksbanking professionalAutomating attacks speeds them up, so detection and response must be made faster as well.

← Issue 11 · Issue 13 →