Issue 2
Viewpoint
oversight · Asia
RBI governor urges banks to speed up AI investment while keeping meaningful human oversight
On 11 August, Sanjay Malhotra, governor of the Reserve Bank of India, urged banks to accelerate AI investment and to keep meaningful human oversight wherever an AI error could seriously harm customers or financial stability. The governor also highlighted the risks.
Who is affected and how
- banking professional
- AI decisions need human review wherever an error could significantly harm customers or stability.
- chief financial officer
- AI investment plans should include skills development and the cost of controls.
case · global
Autonomous AI agents breached a government network, with implications for banks
Reports say a swarm of AI agents breached a government network without human intervention, and the Taiwan case points to a Chinese link. The IMF says AI is raising cyber risks to a new level and…
banking professionalPrepare for automated, AI-driven attacks and speed up vulnerability patching.insurerCyber insurance risks need to be reassessed in light of autonomous attacks.supervision · European UnionEuropean supervisors: banks must defend against cyber threats from frontier AI
The EBA, EIOPA and ESMA issued a joint statement calling for a comprehensive, risk-based and consistent approach to cyber threats from frontier AI. The ECB has also signalled a cyber directive and…
banking professionalReview the IT security plan against AI-based attacks.insurerWith EIOPA involved, insurers also need to assess AI cyber risk.tool · globalOSL, Cloudflare, Orla and Sokin launch stablecoin payment products for AI agents
OSL is entering the AI agent payments market with AgentPay, Cloudflare with programmable wallets, Orla with wallets that have spending caps, and Sokin with an MCP connector. Cloudflare also offers…
fintech companyAgent identification and spending-cap solutions need to be designed for payments made by agents.banking professionalBanks need to assess how payments initiated by AI agents are to be handled on customer accounts.case · United StatesOpenAI CFO outlines AI-native finance function with zero-day close
On 10 August, Sarah Friar, CFO of OpenAI, described lessons from building an AI-native finance function, aimed at a zero-day close and continuous forecasting. She stressed that AI will not replace…
chief financial officerClose and forecasting processes are worth automating, with controls and measured returns, but judgment should stay in human hands.accountantAutomating close and forecasting tasks shifts routine work toward review.Hungary
fraudFraudsters abuse OTP's name with deepfake ads in 'OTP Casino' scamretail customerDo not click or enter data on banking ads or videos seen on social media; verify authenticity through the bank's official channels.banking professionalCustomer service must be ready for reports of such scams and provide proactive customer information.deploymentAI and robots spread quietly through Hungarian banks, with AI complementing robotic automationbanking professionalStaff need to assess which of their processes can be automated and plan AI as an extension of robotic automation, not a replacement for it.chief financial officerBanks' automation and AI development budgets should be planned in line with the robotic tools already in place.fraudDeceptive phone calls on the rise, and fraudsters now target companies tooretail customerNever transfer money or give out data over the phone at anyone's request; call the bank back on its own official number.chief financial officerCompany payment approvals should again include a second verification step independent of the phone.banking professionalFraud warnings and customer education should cover corporate clients as well as retail customers.surveyGránit Bank survey: banking customers increasingly open to an AI assistantretail customerMore banks are likely to offer AI-based spending tracking, so it is worth clarifying what data the system can access.banking professionalCustomer demand justifies building AI-based customer interfaces, with the option of human contact.fintech companyHungarian customers' openness points to a market gap for AI-based personal finance assistants.regulationAI Act: companies using AI may face heavy fines if they fail to meet obligationsfintech companyRisk classification of AI products and compliance documentation should be prepared in advance because of the fine risk.chief financial officerThe cost of AI Act compliance and the fine risk should be built into the budget.regulationHungarian analysis: MNB Recommendation 13/2025 and the AI Act shape AI practice in financebanking professionalThe requirements of the MNB recommendation must also be enforced in supplier and partner contracts.fintech companyAs a bank partner, expect banks to pass on the recommendation's requirements.caseHungarian accountants debate what AI means for their workaccountantRoutine data entry is being automated, so the emphasis should shift to expert review and advisory work.chief financial officerIt is worth agreeing with the accounting firm on its AI use and data-handling practices.
More key stories of the week
case · global
Insurers carve AI damages out of general liability policies
Reports say insurers have quietly introduced exclusions for AI-related damages in general liability policies and are tightening AI coverage. One expert says governance should come ahead of regulation.
insurerDevelop dedicated exclusions and stand-alone products for AI risks, with transparent terms.chief financial officerCheck corporate policies for AI exclusions before a loss occurs.case · United StatesAI chatbots are offering financial advice: how far can they be trusted?
A WBAA report examines whether AI chatbots that give financial advice can be trusted. The stakes in money matters are far higher than with garden or film recommendations.
retail customerA chatbot's financial advice should not be followed unchecked; for larger decisions, use an independent source.banking professionalCustomers increasingly bring chatbot-sourced advice with them, and advisers need to address it in their consultations.case · United StatesAI data center boom is maxing out property and casualty insurers' capacity, AIG CEO says
AIG's CEO says the wave of AI data center construction is "maxing out" the capacity of property and casualty (P/C) insurers. The remarks were published on 13 August.
insurerData center projects create heavy risk concentration, so capacity and reinsurance plans need to be reviewed.chief financial officerInsurance capacity can become a bottleneck when financing data center projects.supervision · globalMoody's: AI race risks "systemic dependency" of big banks on tech firms
Moody's warns that banks' AI race could create "systemic dependency" on Silicon Valley technology companies. The assessment draws attention to big banks' supplier concentration risk.
banking professionalAssess the concentration of critical AI providers and the exit plans for them.regulation · United KingdomFCA's Mills Review sets out vision for AI-enabled retail financial services
According to Legal 500, the Mills Review by the UK Financial Conduct Authority (FCA) charts the direction for retail financial services that use AI. The article was published on 13 August.
banking professionalTrack UK expectations when designing AI-based customer services.fintech companyThe direction of the Mills Review may affect how UK retail AI products reach the market.regulation · United StatesColorado proposes automated decision-making rules; ABA calls for federal AI regulation
Colorado has proposed rules for automated decision-making technology and chatbot safety. The American Bankers Association (ABA) urges federal AI regulation and a level playing field for financial…
banking professionalReview automated decisions and customer chatbots against the new rules.fintech companyFragmented state-level regulation may raise compliance costs.investment · United StatesAI infrastructure financing: shadow credit backstops, data center opposition and a new loan deal
According to Bloomberg, bond traders are worried about roughly $70 billion of shadow credit backstops supporting AI companies, while JPMorgan is leading a $441 million debt deal for an AI…
chief financial officerCheck how transparent the guarantees behind AI infrastructure loans are, and the project risk arising from local opposition.banking professionalFor data center financing, environmental risk and community opposition are part of the credit assessment.fraud · European UnionSpanish police arrest man who used deepfake to bypass video identification
Spanish police arrested a man who used deepfake software to defeat video identification and steal digital signatures for fraud. According to Help Net Security, a split-second glitch exposed the…
banking professionalVideo identification should be strengthened with deepfake detection and additional checks.fintech companyProviders of electronic identification need liveness checks.deployment · AsiaDBS has around 26,000 AI agents running after a few months
DBS let employees build their own AI agents at the end of 2025; within a few months 10,000 were running, then roughly 26,000. Entire banking workflows are being organised around them.
banking professionalAI agents built by employees need a central register and access rules.chief financial officerBuilding workflows on AI agents affects the operating cost structure.supervision · United StatesUS government investigates whether AI investment boom creates economic risk
According to Cyber Daily, the US government is examining whether AI investments are creating economic risk. The article cites the ECB's July concern that if AI overperforms, it will affect financial…
chief financial officerAI-linked exposures are worth testing in stress-test scenarios.In brief
deploymentAI agents for insurers: adesso buys omni:us, NTT DATA and Shift launch new toolsinsurerAI agents in claims handling and underwriting can speed up work, but human review of outputs must be kept.fintech companyAcquisitions in the insurance software market signal consolidation of AI platforms.deploymentGST Council officials study Andhra Pradesh's AI-driven tax administrationaccountantBecause tax authorities use AI analysis, filing data must be consistent and verifiable.chief financial officerTax data quality and the reconcilability of accounts matter more as official checks become automated.deploymentMiddle Eastern and Indian banks build on AI-native, agentic modelsbanking professionalAI-native banking models represent competition, so experience with deploying agentic services is worth following.fintech companyBanks' AI and fintech partnerships open new opportunities for cooperation.deploymentIndian banks move from AI experimentation to deployment at scalebanking professionalPilot AI projects must be made scalable across lending, productivity and risk.chief financial officerScaling costs and returns should be measured separately from pilots.investmentHedge funds rotated from AI infrastructure to AI monetisation plays in Julychief financial officerPositioning in AI-linked investments is shifting, which warrants a review of portfolio risk.regulationBermuda, BIS, Slovenia and Poland weigh in on AI rules for financial servicesinsurerThe Slovenian sandbox signals that a regulatory route is opening for controlled testing of insurance AI.caseEY uses AI in audit and appears likely to become Anthropic's auditoraccountantThe use and documentation of AI tools in audit should be set up with PCAOB inspections in mind.chief financial officerThe overlap between auditor relationships and AI providers calls for an independent assessment.fraudAI-enabled fraud raises risks for banks in South Africa and the Philippinesbanking professionalMulti-layered verification is needed to counter attempts to bypass deepfake-based identity checks.fraudDeepfake Ratan Tata investment scam in India; Guardian reports on AI scam exploiting holiday photosretail customerA video investment offer from a well-known person should not be accepted without verification, and photos shared on social media can serve as the basis for targeted scams.banking professionalCustomer warnings should be extended to investment scams that use deepfake endorsements.toolAI fraud-detection tools from Yiren Digital, KFin and EDGE–Socure target lossesbanking professionalTesting fraud-detection and signature-verification AI tools is warranted to reduce losses.fintech companyIn lending fraud, AI-based identification and verification may become a market requirement.oversightSEBI uses AI to monitor social media for fraudretail customerFor investment offers on social media, check that the provider is licensed.caseAI credit scoring: first-time borrowers in Egypt and legal gray areas across bordersbanking professionalFor AI credit scoring, data sources and cross-border legal requirements need to be clarified.retail customerFor customers without a credit history, an AI model may open up new access to credit.toolMercury launches Spend with agent cards and intelligent budgets for AI agentschief financial officerCards issued to AI agents need spending limits and approval rules.fintech companyThe agent card signals a new product category in spend management.fraudASIC reports record removals of AI-generated fake celebrity investment adsretail customerAn investment endorsed by a celebrity should not be accepted without checking the licence.deploymentAI-powered insurance shopping rivals brokers; Suncorp and IAG deepen AI useinsurerTo appear in AI channels, product and quote data must be made machine-readable.retail customerWhen comparing insurance through an AI assistant, check the range of offers being presented.chief financial officerBudget planning for AI investment in the sector is growing significantly.caseAustralia's TPB issues AI guidance as Germany's HSP rolls out ChatGPT in tax advisoryaccountantAI use should be documented and supervised in line with professional body expectations.chief financial officerAdvisory providers should be asked about their AI use, particularly on data protection and liability.layoffsVisa cuts 2,600 jobs as AI reshapes its businessfintech companyAI-driven restructuring may continue across the payments sector, so organisational plans should be aligned with it.banking professionalThe transformation of the payments network may affect partner relationships and service levels.surveyWorkiva survey: five lessons on AI adoption in financial serviceschief financial officerBenchmark data lets finance leaders compare their own AI maturity and governance practices with the market.caseJPMorgan Chase chief leads coalition on responding to increasingly powerful AIbanking professionalThe big-bank initiative may signal sector directions for managing AI risks.surveyTSB research: more than half of those following social media financial advice lost moneyretail customerFollowing social media or AI-based investment advice without checking it risks losses.banking professionalIt is worth providing young customers with credible, verifiable investment information.