A widening divergence in monetary policy alongside rapid advances in artificial intelligence is shaping how risks are assessed and regulated across the banking sector. Central banks are increasingly navigating different inflation narratives and policy constraints, while regulators and institutions are simultaneously adapting to the systemic implications of AI adoption in financial services. In the United States (US), Federal Reserve vice chair Michelle Bowman signalled that policymakers would look through energy-driven inflation, warning that a tightening response could place unnecessary pressure on economic activity and labour markets. Bank of Japan (BOJ) Governor Kazuo Ueda delivered an academic address in Tokyo contextualising the current oil shock against five decades of inflation history, signalling that the BOJ remains alert to the risk of persistence and keeping a June rate hike firmly on the table. US growth data also pointed to softer economic momentum alongside persistent inflation. At the same time, banks and regulators accelerated work on artificial intelligence (AI). TD Bank deployed an agentic AI model in mortgage and home equity lending, Robinhood granted AI agents direct access to customer trading and payment accounts, and the International Organization of Securities Commissions (IOSCO) released a supervisory toolkit for AI use in capital markets. Regulators in Australia, Hong Kong and the United Kingdom (UK) also advanced new frameworks covering prudential proportionality, operational resilience and AI-enabled cyber risk. The European Central Bank (ECB), in its May Financial Stability Review, highlighted vulnerabilities in the non-bank financial sector as the system’s most persistent unresolved risk. Read more on the week's key developments: 1. Fed vice chair Bowman signals no rate hike despite rising inflation Federal Reserve vice chair for supervision Michelle Bowman said on 29 May that policymakers should look through the recent rise in energy-driven inflation, arguing that tightening policy in response would place unnecessary pressure on economic activity and labour markets. Her remarks at the Reykjavík Economic Conference hosted by the Central Bank of Iceland followed April personal consumption expenditures (PCE) inflation of 3.8% — the highest reading since mid-2023, driven by Iran war energy prices — and core PCE at 3.3%. She supported keeping the existing Federal Open Market Committee (FOMC) statement language that leaves room for further rate cuts, directly separating herself from a growing faction of policymakers willing to signal that the next move could be a hike. The speech is the first substantive public policy signal from within the leadership of Chair Kevin Warsh's Federal Reserve, sworn in on 22 May. Warsh himself has not yet spoken publicly on rates. The funds rate remains at 3.50%–3.75%, held since the 28–29 April FOMC meeting. Markets are pricing no cuts through end-2026 and a small probability of a hike in early 2027. 2. US growth slows as inflation remains elevated Revised US economic data showed first-quarter gross domestic product (GDP growth) of 1.6% annualised, below earlier estimates of around 2%, while inflation remained above the Federal Reserve's target, according to the Bureau of Economic Analysis. The American Bankers Association noted on 27 May that the combination of slower growth and persistent price pressures complicates the policy outlook and increases the likelihood that banks will continue to operate in an environment of elevated funding costs and rising credit-risk concerns. The GDP miss, set against still-sticky inflation, points to a stagflationary configuration that complicates both monetary policy and bank credit modelling. Slower growth raises borrower stress and widens expected credit losses, while persistent inflation limits the Fed’s scope to ease, keeping funding costs elevated and deposit repricing pressure in place. For banks, this combination is particularly adverse because it weakens both sides of the income statement at once: credit quality deteriorates while net interest income (NII) remains constrained. Stress scenarios increasingly need to incorporate a prolonged period of sub-trend growth alongside above-target inflation — a regime in which asset quality and margins erode concurrently rather than cyclically offset each other. 3. BOJ governor Ueda warns oil shocks can become persistent Bank of Japan Governor Kazuo Ueda told participants at the BOJ-Institute for Monetary and Economic Studies (IMES) Conference on 27 May that the impact of oil shocks depends on prevailing wage and inflation dynamics, drawing comparisons with previous episodes in Japan's inflation history. The remarks come ahead of the BOJ's June policy meeting, with markets increasingly expecting a further rate increase after inflation remained above target and the April decision produced a 6-3 split vote. The speech reinforces expectations that the BOJ remains alert to inflation persistence and supports the case for further policy normalisation, a favourable backdrop for Japanese bank earnings. The BOJ held its policy rate at 1.0% at its April 28 meeting — a split 6-3 vote, with dissenters arguing for a hike — and the June meeting is now the focus of hike expectations, with inflation running at 2.8% on a core basis. Japanese megabanks MUFG, SMFG, and Mizuho each flagged the Iran conflict as a downside earnings risk at their May results briefings. The divergence between Japanese banks — which gain from rate hikes — and their regional peers facing net interest margin (NIM) compression makes the BOJ trajectory the single most important rate variable for Asian banking sector earnings in the second half of 2026. 4. ECB highlights non-bank financial vulnerabilities in Financial Stability Review The European Central Bank's (ECB’s) May 2026 Financial Stability Review, published 27 May, identified persistent liquidity and leverage risks in the non-bank financial intermediation (NBFI) sector, calling for stronger supervisory coordination, improved data collection and new macroprudential tools to address vulnerabilities in private markets and investment funds. The review argues that existing regulatory frameworks remain insufficient to address systemic risks outside the banking sector, increasing the importance of banks' own risk management of exposures to non-bank counterparties. The ECB's review forces banks to reckon with a structural gap: the NBFI sector, which includes money market funds, open-ended funds, private credit platforms, and other non-bank lenders, operates at a scale that can amplify system-wide stress, yet lacks the macroprudential toolkit that applies to banks. European banks' interconnections with NBFI are substantial and growing. A liquidity event in the NBFI sector, triggered by energy market dislocation, a credit cycle turn, or geopolitical shock, transmits directly to bank balance sheets through repo, prime brokerage, credit lines, and fund manager relationships. Banks cannot assume that ECB or national supervisory action will contain NBFI stress in real time; they must manage their own NBFI exposure as a primary risk, not a secondary one. 5. UK authorities warn frontier AI could enable synchronised cyberattacks The Bank of England, Financial Conduct Authority and His Majesty’s Treasury warned on 15 May that advanced AI models could identify software vulnerabilities and enable coordinated cyberattacks across financial institutions at a scale not previously possible. The warning elevates AI-enabled cyber risk from an operational concern to a supervisory priority and is likely to drive further investment in AI-specific resilience and threat-monitoring capabilities across the banking sector. Frontier AI cyber risk is no longer a theoretical concern — it is now a regulatory supervisory priority in the UK, with other major jurisdictions expected to follow. Banks that have concentrated AI governance frameworks on deployment risk (bias, explainability, model drift) must now extend those frameworks to adversarial risk: the use of AI by external actors to exploit bank systems. The immediate implication is a repriorisation of cybersecurity investment toward AI-specific defences: automated threat detection, real-time vulnerability scanning and continuous red-team testing using adversarial AI tools. The reputational and operational cost of a synchronised AI-enabled attack across multiple institutions simultaneously would be categorically more severe than past cyber incidents. 6. Hong Kong operational resilience deadline takes effect The Hong Kong Monetary Authority (HKMA) implementation deadline for its Operational Resilience framework (Supervisory Policy Manual OR-2) took effect on 31 May, requiring all authorised institutions to demonstrate the ability to maintain critical operations during severe disruptions. The milestone completes a multi-year transition programme and raises supervisory expectations around banks' preparedness for cyber incidents, geopolitical shocks and third-party technology failures. Banks operating in Hong Kong that have not completed the framework implementation are now exposed to regulatory action. The OR-2 standard requires banks to map and test every critical operation — payments, custody, trade finance, lending and IT infrastructure — against disruption scenarios, which must include geopolitical shock , cyberattack and third-party technology provider failure. For international banks with Hong Kong operations, the 31 May deadline means OR frameworks are now a standing supervisory deliverable. Failure to demonstrate compliance risks remediation orders and, in severe cases, activity restrictions. The HKMA's OR-2 framework is also a regional benchmark: Singapore's MAS, the Reserve Bank of Australia, and the Reserve Bank of India have each published aligned operational resilience guidance, meaning banks with multi-market Asian operations must run comparable programmes across all jurisdictions simultaneously. 7. Canada’s TD Bank Group deploys agentic AI in mortgage lending TD Bank Group launched an agentic AI model to support mortgage and home equity line of credit applications, marking one of the first publicly disclosed deployments of autonomous AI within a core retail lending process. Unlike generative AI (GenAI) copilots that assist human operators, agentic AI models execute multi-step tasks autonomously — in this case, moving applicants through the mortgage process without requiring manual intervention at each stage. TD's mortgage deployment marks the shift from AI as productivity tool to AI as operational infrastructure in front-line lending. The competitive implications are significant: banks with deployed agentic systems can process applications faster, at lower cost, with fewer errors and less staff — a structural cost advantage in the highest-volume retail lending categories. Banks that remain in pilot phases or restrict AI to back-office functions face an expanding gap in unit economics. Regulators will now be required to address agentic AI in lending under fair lending, explainability, and consumer protection frameworks — a supervisory gap that does not yet have clear resolution in the US, European Union or UK. The first banks to deploy also absorb the regulatory uncertainty risk of being the test cases. 8. IOSCO releases AI supervisory toolkit for capital markets The International Organization of Securities Commissions (IOSCO) published on 25 May a Supervisory Toolkit for AI Use in Capital Markets, providing regulators with a framework for overseeing AI systems across their lifecycle, including emerging agentic AI applications. The framework is a product of IOSCO's Fintech Task Force, whose membership spans over 200 regulators across 130 jurisdictions. It provides supervisors with a three-layer structure: first, areas of supervisory consideration for AI systems in capital markets; second, supervisory tools and approaches for each area; and third, indicators and data sources for monitoring AI adoption, including on-site inspection frameworks, targeted survey mechanisms and documentation requests. The framework is non-binding and non-prescriptive, designed to be applicable across different national regulatory models. The IOSCO toolkit is a globally coordinated supervisory framework specifically designed for AI oversight in capital markets, and it arrives at the moment when AI deployment is accelerating most rapidly across trading, risk, compliance, and client-facing functions. For banks with capital markets operations, which encompasses virtually every major institution with investment banking or asset management arms — the toolkit signals the beginning of structured, cross-border regulatory examination of AI systems. 9. US fintech firm Robinhood launches agentic trading and payments US fintech company Robinhood launched Agentic Trading and an Agentic Credit Card on 27 May, allowing third-party AI agents to execute trades and make purchases on behalf of its 27 million customers within predefined limits. Using the open Model Context Protocol, customers can connect third-party AI assistants — including those from OpenAI and other platforms — to a dedicated, isolated trading account and a virtual credit card with user-controlled spending limits. The AI agent can execute trading strategies, rebalance portfolios and make retail purchases autonomously, without requiring human confirmation at each step. The launch represents one of the most significant retail finance applications of agentic AI to date and could accelerate the shift of customer engagement from traditional banking channels to AI-driven interfaces. When AI agents hold direct access to trading accounts and payment instruments and can act without step-by-step human approval, banks lose the transactional moment at which they provide value, advice, and data. The customer relationship migrates upward to the AI layer, which becomes the interface for savings allocation, investment execution, and spending. Banks that do not deploy their own agentic interfaces risk becoming infrastructure providers, processing transactions they no longer see for customers they can no longer engage. 10. Australia formalises three-tier banking framework The Australian Prudential Regulation Authority confirmed on 27 May that a new three-tier framework for banking supervision will take effect on 1 July 2026, differentiating prudential requirements according to institution size and complexity. Under the new framework, banks are classified into three tiers: Most Significant Financial Institutions (MSFIs), defined as banks with total assets above AUD 300 billion ($195 billion), which will face the highest prudential requirements; Significant Financial Institutions (SFIs), with the SFI threshold raised from AUD 20 billion ($13 billion) to AUD 30 billion ($19 billion); and non-SFIs, comprising all remaining banks, which will be given additional time to comply with new or revised prudential requirements relative to the other tiers. The reforms are intended to improve proportionality within the regulatory framework and may reduce compliance burdens for smaller institutions while maintaining stricter requirements for the largest banks. For the large banks — ANZ, Commonwealth Bank of Australia, National Australia Bank, and Westpac — MSFI classification means they will carry the heaviest capital, liquidity, and operational resilience requirements and will serve as the reference point for the most stringent tier. For mid-tier and regional banks, the raised SFI threshold and explicit compliance lead-time commitments reduce regulatory drag and may improve capital efficiency and product pricing competitiveness. The framework also creates a template that other regulators in the Asia-Pacific region, particularly those managing similarly concentrated banking systems, are likely to study and adapt What to watch: US Federal Open Market Committee meeting (17–18 June); Bank of Japan Monetary Policy Meeting (15–16 June); European Central Bank rate decision (10—11 June); HSBC interim results (4 August); Goldman Sachs European Financials Conference (3 June) The Asian Banker Weekly Brief is a roundup of the biggest macroeconomic, industry and regulatory developments affecting banking globally.
