Current thematic dossiers

Select a theme to view its dossier.

Cyber & operational resilience

This deep dive takes stock of recent developments in cyber and operational resilience, including current supervisory risk watch points, the direction of policy and guidance as well as evolving supervisory practices.

Overview

Recent supervisory insights point to six recurring operational resilience and ICT concerns:

  • Operational and ICT risk controls that are not consistently executed, evidenced or independently challenged
  • Insufficient independent assurance over critical ICT providers and intra-group services
  • Critical-service mapping and impact tolerances that do not consistently drive severe end-to-end testing
  • Change and migration controls that are not preventing avoidable service disruption
  • Uneven fundamental cyber controls and visibility over the technology estate
  • Incident classification, reporting and remediation that do not consistently support effective learning

Many of these watch points now reflect the first wave of DORA-specific supervisory assessment activity. From a supervisory perspective, the focus is whether firms can demonstrate that DORA implementation has moved beyond formal project plans, policies and gap analysis into reliable operating practice: complete and well-documented ICT risk-management frameworks, accurate registers of information, effective contract, subcontracting and oversight arrangements for ICT third-party providers, business-impact analysis for critical or important functions, and reliable classification and reporting of major ICT incidents among other things.

Taken together, the DORA assessment findings and broader operational-resilience, ICT risk and cyber hygiene observations point to the same practical concern: firms may have frameworks in place, but they do not always execute them consistently, evidence them clearly, challenge them independently, remediate weaknesses promptly or translate incidents and tests into learning points that drive concrete oversight and control changes.

What's new

HM Treasury has added five suppliers to the Critical Third Parties regime, increasing the total number designated to nine. The Bank of England, Prudential Regulation Authority and Financial Conduct Authority will jointly oversee the suppliers’ specific systemic services to the financial sector, with oversight limited to those services rather than their wider operations.

The Reserve Bank of New Zealand has published its Financial Stability Report, assessing banks and insurers as capable of absorbing severe shocks despite growing geopolitical, market and cyber risks. Domestic financial stress has declined unevenly, with smaller businesses and sectors reliant on discretionary spending still under pressure. Financial institutions need to strengthen cyber defences and service recovery capabilities as frontier AI accelerates attacks and shared technology providers increase the potential for disruption across the system.

The European Central Bank has published an analysis warning that advances in artificial intelligence could overwhelm cyber defences and increase systemic risk for euro area banks. Unequal access to advanced models and reliance on a small number of technology providers compound these vulnerabilities. The analysis calls for stronger controls over AI, coordinated contingency planning and testing, and mapping of shared technology exposures.

Deep dive

Frontier AI is shortening vulnerability, patching and response windows

Supervisors increasingly treat frontier AI as a strategic operational resilience issue stressing that advanced models can increase the speed, scale, accessibility and sophistication of vulnerability discovery, exploitation and attack execution. The concern is that weaknesses in firms, market infrastructures and shared digital infrastructure may be exploited before defenders can respond, potentially turning isolated vulnerabilities into wider operational disruption.

Anthropic’s April 7 release of Claude Mythos Preview was followed by a concentrated series of financial-sector warnings in the second quarter of 2026. The model was reported to have autonomously identified thousands of previously unknown vulnerabilities in widely used operating systems and web browsers, while a UK AI Security Institute assessment found that it was the first model to autonomously sequence a complete 32-step corporate-network attack, completing all steps in three of 10 attempts. Authorities broadly warned that such capabilities could make sophisticated attacks faster, cheaper and accessible to less skilled actors, narrowing the interval between vulnerability discovery, exploitation and remediation and increasing the risk of simultaneous failures across common software, cloud, payments and data infrastructure.

In Australia, the Australian Prudential Regulation Authority said information security practices were struggling to keep pace and flagged prompt injection, data leakage, insecure integrations, exploit injection, autonomous agent misuse, incomplete security testing and patching timelines that did not match the accelerated threat. The Australian Securities and Investments Commission warned that isolated weaknesses could have a systemwide domino effect. The Securities and Exchange Board of India cited the potential for cascading effects across the interconnected securities market and established the cyber-suraksha.ai task force, while the Financial Conduct Authority, Bank of England and HM Treasury said current frontier models already exceed what a skilled practitioner could achieve at greater speed, larger scale and lower cost. Japan’s Financial Services Agency and the Bank of Japan called for short-term preparations for the rapid discovery of large numbers of vulnerabilities and the resulting volume of patches, while South Korea’s Financial Services Commission announced arrangements to support AI-driven cyber defense. Across these actions, authorities reinforced expectations that boards and senior management treat frontier AI as a management and operational-resilience issue, accelerate vulnerability triage and patching, strengthen access, network and data controls, monitor third-party and open-source dependencies and maintain tested containment, incident response, recovery and business continuity arrangements.

Following the initial wave of warnings, authorities have more formally accounted for the impact of frontier AI in their risk assessments. The European Systemic Risk Board (ESRB) issued a warning in July on systemic cyber risks from frontier AI models, after its General Board had raised its assessment of systemic cyber risk from elevated to severe in June. The European Insurance and Occupational Pensions Authority (EIOPA) moved digitalisation and cyber risk to a high level in its July risk dashboard for insurers, attributing the change to frontier AI developments alongside geopolitical tensions and cyber underwriting exposures, and the European Securities and Markets Authority (ESMA) reported in September that its outlook for operational risk was rising as frontier AI changes the cyber threat landscape. Evidence of the shortening window between discovery and exploitation has since started to accumulate. The Monetary Authority of Singapore (MAS) reported that high-severity vulnerabilities rose sixfold in 2026 to 2,200 against the preceding three-year average, and the International Monetary Fund found that in more than 70% of the cases it examined this year, attackers exploited software flaws before providers could patch them.

In parallel, supervisors have further formalised their expectations towards entities. At EU level, ECB Banking Supervision directed significant institutions to assess how AI-enabled threats affect their risk profile and to submit action plans by 31 October, covering exposed assets, accelerated patching, monitoring and third-party risk. The European Supervisory Authorities on their part have begun engaging the critical ICT third-party providers under their oversight on AI-related risks, which will feature in their 2027 examinations. Sweden’s Finansinspektionen opened on-site reviews in September of how firms prevent, detect and manage AI-enabled attacks. In Singapore, MAS made AI-assisted red teaming of critical internet-facing systems mandatory for key institutions from 1 July and will require risk assessments and action plans.

Firms lack sufficient visibility, contractual leverage and independent assurance over critical ICT providers

Supervisors are concerned that firms remain heavily dependent on external or group service providers without maintaining an adequate independent basis for assessing the resulting risks. Notably, oversight of critical external and intra-group ICT services does not always provide an independent and sufficiently detailed view of provider performance, access, subcontracting, control effectiveness and continuity.

European Central Bank (ECB) Banking Supervision has identified that some banks remain behind in renegotiating third-party contracts and establishing adequate business continuity arrangements, and has made closure of these gaps a focus of an ICT third-party-risk on-site campaign. Across ICT inspections of banks and insurers, the Financial Supervisory Authority of Norway (Finanstilsynet) found that firms sometimes relied too heavily on assurance supplied by providers, groups or industry alliances. The authority expected firms’ control functions to maintain their own documented basis for assessing provider risk, service-level performance, access rights and compliance, and to conduct targeted checks where the criticality or complexity of a service required them, and has since described ICT third-party risk management as still maturing, citing complex supplier chains as a particular difficulty. Reviews by other authorities point in the same direction. The Danish Financial Supervisory Authority found that some insurance and pension companies still lack a complete overview of provider-dependent processes and called for stronger board reporting on material provider developments. The Dutch Authority for the Financial Markets (AFM) found that large asset managers did not always write their recovery objectives into agreements with ICT providers, and the Canadian Securities Administrators found that 62% of the 73 registered firms it reviewed had little or no documentation of how they oversee their technology providers. Securing adequate contractual terms also remains a practical obstacle: France’s Autorité des marchés financiers (AMF) reported that some providers had refused to be identified as ICT third-party providers, disputed the need for contractual remediation or charged additional fees for audit and testing rights.

The concerns are reinforced by incident data. Around one-third of the 3,383 major incidents reported by EU financial entities for 2025 had a cross-border impact according to the European Supervisory Authorities (ESAs)’ analysis, underscoring how shared infrastructure, services and dependencies can transmit disruption across firms and jurisdictions. At national level, the Bank of Italy found that a service provider or subcontractor was involved in 56% of the 137 major incident notifications it received for 2025, with almost all operational incidents at less significant institutions originating at providers, and the AMF attributed 27 of the 31 confirmed major incidents at portfolio management companies to providers.

Fundamental cyber controls and visibility over the technology estate remain uneven

Supervisors continue to identify weaknesses in asset and dependency inventories, privileged access management, secure configuration, vulnerability detection, security monitoring, logging, backup testing and restoration preparedness. The concern is the coverage and operating effectiveness of these controls, particularly across externally managed systems.

In a thematic review of regulated markets and other trading platforms, the Dutch Authority for the Financial Markets found that elements of the ICT risk-management framework required stronger coverage, particularly security monitoring, access management, logging, emergency change controls and continuity management. Its separate SREP work also identified weaknesses in vulnerability detection, backup testing and incident preparedness. Cross-sector examination findings published by the Taiwan Financial Supervisory Commission (FSC) showed comparable deficiencies across financial institutions, including incomplete controls and audit trails for the external transmission of personal data, inappropriate use or administration of privileged accounts, and inadequate host-security baselines and configuration standards. In South Korea, a review of 555 firms by the Financial Services Commission following a series of intrusions found deficiencies at 130, mainly missing security patches, unsupported systems, a lack of continuous monitoring and inadequate controls over externally exposed systems. The common concern is that firms may possess overarching cyber frameworks while lacking complete visibility of the assets, access rights and configurations to which those frameworks must be applied. Enforcement cases in Australia, Hong Kong and South Korea in 2026 illustrate the consequences, with sanctions imposed over weaknesses such as unresolved penetration-test findings, unpatched systems, weak access controls and inadequate backups.

Change and migration controls are not preventing avoidable service disruption

Deficiencies in change classification, risk assessment, testing, approval, rollback and post-implementation verification are allowing technology changes and migrations to cause material operational incidents. Supervisors are particularly attentive to major transformations, emergency changes and high-volume or pre-approved changes.

ECB Banking Supervision reported that IT change was identified as the root cause of 38% of major incidents reported by banks for 2025 and has designated change management processes and controls as a specific supervisory focus. Following a March 2026 disruption at a major Singapore retail bank, the Monetary Authority of Singapore (MAS) found that an erroneous step during a system change had interrupted digital banking services and stated that it would follow up on the bank’s change-management process.

Critical service mapping and impact tolerances do not consistently drive severe end-to-end testing

Firms cannot always demonstrate that critical or important services will remain within tolerable disruption levels. Mapping and business impact analysis are not consistently translated into recovery priorities, test scenarios, fallback arrangements and evidence of end-to-end recoverability.

In its review of operational resilience self-assessments, the UK Financial Conduct Authority found that some firms’ mapping remained overly focused on technology and did not adequately capture people, processes, facilities, information and third-party dependencies. It also found cases where firms had not distinguished between tolerances for consumer harm and market integrity effects, claimed that they could recover from any disruption without sufficiently severe testing, or had not tested alternatives where normal communication channels were unavailable. Across inspections of banks and insurers, Norway's Finanstilsynet similarly found that business-impact analyses could be too high-level or insufficiently connected to availability requirements, control priorities and testing. It called for firm-level tests covering complete business processes, prolonged outages, serious cyber incidents, provider failures and worst-case scenarios.

Operational and ICT risk controls are not consistently executed, evidenced or independently challenged

Firms do not always translate operational and ICT risk frameworks into routinely performed controls, clear ownership, effective independent challenge and timely remediation. Board oversight is also weakened where approval trails, management information and accountability for corrective actions are unclear.

As part of a SREP assessment covering investment institutions, investment firms and trading and settlement platforms, the Dutch Authority for the Financial Markets (AFM) found that internal controls were not always performed consistently or properly documented, processes were not reviewed regularly, and responsibilities for ownership, oversight and monitoring were insufficiently clear. Similar patterns also emerged in other jurisdictions: The Financial Conduct Authority (FCA) identified in its review of operational resilience self-assessments across banks, insurers, payment firms, investment firms and market infrastructures: some firms could not evidence clear board approval, named owners and completion dates for remediation, or meaningful input from the second and third lines of defence. A series of ICT inspection reports issued by Norway’s Financial Supervisory Authority (Finanstilsynet) in 2026 covering banks and an insurer highlighted limited operationalisation of board-approved governance or steering documents, insufficient ICT capacity or expertise in parts of the control environment, gaps in ICT-risk reporting and control documentation, and, in at least one bank inspection, limited follow-up of internal-audit recommendations. In its review of 14 insurance and pension companies, the Danish Financial Supervisory Authority identified weaknesses in board oversight of ICT risk, unclear allocation of responsibilities and insufficient separation between operational and control functions, alongside gaps in documentation and staff capabilities.

Incident classification, reporting and remediation do not consistently support effective learning

Incident severity assessment, root-cause analysis, loss measurement, regulatory reporting and remediation tracking remain inconsistent. This reduces the reliability of supervisory information and increases the risk that recurring control failures are detected without being prevented.

In a SPOT inspection campaign covering five portfolio-management companies and the period from 2022 to 2024, France’s Autorité des marchés financiers (AMF) found that the severity of operational incidents and near misses was not always assessed systematically, remediation plans were only partially tracked, and the materiality criteria used to exclude incidents from management reporting were not consistently documented. The AMF also identified significant inaccuracies in operational-loss data reported to the authority by one firm and found that some firms maintained multiple, non-harmonised incident registers. At EU level, the ESAs’ first annual analysis of major incidents reported under the Digital Operational Resilience Act identified divergent reporting practices across sectors and jurisdictions. Although partly attributable to the framework’s first year of operation, those differences limit data quality, comparability and supervisory usability and reinforce the need for consistent classification, complete root-cause reporting and verifiable remediation closure.

Key sources