Overview
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.
State of play
Operational resilience is the ability of financial institutions, and of the system they form, to maintain or restore critical services through disruption. Holistic operational resilience frameworks are now in force across the major geographies: the EU’s Digital Operational Resilience Act has applied since January 2025, the UK regime completed its transition in March 2025, APRA’s prudential standard on operational risk management (CPS 230) took effect in Australia in July 2025, and Hong Kong, Canada and India followed in 2026. With these frameworks in place, the supervisory attention is now shifting towards the effectiveness of these frameworks in practice. The first year of harmonised incident data in the EU offered a preliminary market-wide reading: EU financial entities reported 3,383 major ICT incidents in 2025. System failures and external events were the main drivers and around one third of the incidents involved cross-border impact. Additional supervisory analyses point to the same two sources of disruption: IT change, which ECB Banking Supervision identified as the root cause of 38% of major incidents at significant banks, and third parties, which accounted for 27% of incidents reported to the UK FCA and caused 87% of major incidents at French portfolio managers. Findings elsewhere point the same way: APRA notes that large Australian institutions each rely on around 150 material service providers, and South Korea’s Financial Services Commission found deficiencies,…
View allView less
Operational resilience is the ability of financial institutions, and of the system they form, to maintain or restore critical services through disruption. Holistic operational resilience frameworks are now in force across the major geographies: the EU’s Digital Operational Resilience Act has applied since January 2025, the UK regime completed its transition in March 2025, APRA’s prudential standard on operational risk management (CPS 230) took effect in Australia in July 2025, and Hong Kong, Canada and India followed in 2026. With these frameworks in place, the supervisory attention is now shifting towards the effectiveness of these frameworks in practice. The first year of harmonised incident data in the EU offered a preliminary market-wide reading: EU financial entities reported 3,383 major ICT incidents in 2025. System failures and external events were the main drivers and around one third of the incidents involved cross-border impact. Additional supervisory analyses point to the same two sources of disruption: IT change, which ECB Banking Supervision identified as the root cause of 38% of major incidents at significant banks, and third parties, which accounted for 27% of incidents reported to the UK FCA and caused 87% of major incidents at French portfolio managers. Findings elsewhere point the same way: APRA notes that large Australian institutions each rely on around 150 material service providers, and South Korea’s Financial Services Commission found deficiencies, chiefly unpatched and unsupported systems, at 130 of the 555 firms it reviewed in October 2026.
Two developments have materially impacted the risk picture in 2026. The first is frontier AI. After Anthropic in April released Claude Mythos Preview, a model reported to discover and exploit vulnerabilities autonomously, supervisors across Europe, Asia-Pacific and North America issued a concentrated series of warnings in the second quarter, raised their assessments of systemic cyber risk and set up dedicated task forces and working groups, while the IMF, the FSB and the BIS reinforced the message at global level. The concern is less a new form of attack than a compression of time: frontier models lower the time, cost and expertise needed to find and exploit weaknesses, narrowing the window between a vulnerability’s discovery and its exploitation, while defenders remain slower because fixes must be tested before they are deployed. The second is concentration. Institutions depend on the same cloud, software and AI providers, so that a weakness in one can become a simultaneous failure across many.
The policy and supervisory response has converged on four lines. On frontier AI, authorities are intensifying existing frameworks rather than creating AI-specific regimes. Boards are expected to own the risk, patching to be accelerated and recovery to be tested against faster attacks. Several supervisors, including ECB Banking Supervision and the Monetary Authority of Singapore, are now requiring action plans. On third parties, oversight is extending from firms’ outsourcing controls to the providers themselves. The ESAs are examining the 19 critical ICT providers designated in 2025, the UK brought four cloud providers under direct oversight in July 2026, and CPMI-IOSCO is consulting on whether the growing reliance of market infrastructures on external providers calls for further policy work. On incidents, reporting is being harmonised around the FSB’s Format for Incident Reporting Exchange, a common template for reporting operational and cyber incidents published in April 2025, so that the same event is classified consistently across sectors and borders. On preparedness, testing is becoming more demanding. Sector-wide exercises now simulate shared dependencies: in October the Bank of England ran a sector-wide exercise in which the authorities and 38 of the largest banks and market infrastructures rehearsed their coordinated response to a global outage of cloud services. Cyber events are also being built into conventional financial stress tests alongside credit and market shocks, as in the joint APRA–RBNZ stress test of October 2026, which included a cyberattack on a bank’s offshore operation. Firm-level testing is becoming more technical, with threat-led penetration testing under DORA and AI-assisted red teaming required in Singapore.
Recent developments
(Last update: )
Warnings over frontier AI have now translated into more concrete supervisory actions. ECB-supervised banks must submit their AI cyber-risk action plans by 31 October. Sweden’s Finansinspektionen has opened on-site reviews of how firms prevent, detect and manage AI-enabled attacks. Thailand’s Office of Insurance Commission ran two sector-wide cyber exercises with 67 insurers in late September, built around AI-related threats and failures at external providers.
New supervisory findings point to some of the practical challenges and remaining deficiencies in firms’ management of third-party dependencies. Visibility is incomplete: the Danish FSA found that some insurers and pension companies still lack a full overview of provider-dependent processes, and around 80% of banks surveyed by the EBA cite dependence on ICT providers outside the EU/EEA as a challenge. Contract terms are contested: French asset managers reported providers refusing to be classified as ICT third-party providers or…
View allView less
Warnings over frontier AI have now translated into more concrete supervisory actions. ECB-supervised banks must submit their AI cyber-risk action plans by 31 October. Sweden’s Finansinspektionen has opened on-site reviews of how firms prevent, detect and manage AI-enabled attacks. Thailand’s Office of Insurance Commission ran two sector-wide cyber exercises with 67 insurers in late September, built around AI-related threats and failures at external providers.
New supervisory findings point to some of the practical challenges and remaining deficiencies in firms’ management of third-party dependencies. Visibility is incomplete: the Danish FSA found that some insurers and pension companies still lack a full overview of provider-dependent processes, and around 80% of banks surveyed by the EBA cite dependence on ICT providers outside the EU/EEA as a challenge. Contract terms are contested: French asset managers reported providers refusing to be classified as ICT third-party providers or charging additional fees for audit and testing rights. Recovery commitments are not aligned with the business functions providers support, as the Dutch AFM found at large asset managers. Similar challenges are observed at market infrastructures: a new CPMI-IOSCO survey, released in September, highlights that around a third of infrastructures have visibility only as far as their direct providers, and that around three quarters rate the risk from provider concentration as high or medium.
Concerns continue to be echoed at the global level: In its latest Global Financial Stability Report, released against the backdrop of the Annual Meetings in Bangkok in October, the International Monetary Fund names systemic cyber threats as one of three near-term triggers for market turmoil. Its assessment rests on the speed at which weaknesses are now being exploited: in more than 70% of the cases it examined this year, attackers used software flaws before providers had been able to patch them. The FSB chair’s letter to G20 finance ministers ahead of their 15 October meeting confirms that AI-related cyber risk will remain on the FSB’s work programme in 2027, building on his August call for jurisdictions to support safe and responsible releases of frontier models and for institutions to secure the resilience of their critical technology providers.
Timeline of developments
Bank of Italy highlights operational resilience gaps and technology provider risks in Senate testimony
Bank of Italy financial supervision and regulation head Giuseppe Siani told a Senate inquiry that DORA implementation has improved, but gaps remain in technology controls and supplier oversight. Nearly 60% of incident reports in the first nine months of 2026 were attributable to third parties. He also recalled the end of 2026 deadline for directly supervised intermediaries to prepare cybersecurity and resilience plans addressing advanced artificial intelligence risks.
South Korea's Financial Services Commission details cybersecurity checks at about 580 firms and plans stronger AI defenses
South Korea's Financial Services Commission has detailed security checks at about 580 firms and plans for at least two vulnerability assessments annually, starting with large firms. Fraud safeguards include an artificial intelligence phishing risk score and proposed partial compensation for voice phishing losses regardless of fault. Further plans would replace network separation requirements with alternative safeguards and extend security regulation through a Digital Financial Safety Act.
Angola’s Capital Markets Commission chair calls for flexible regulation of artificial intelligence and cloud computing
At the Africa Financial Summit 2026, Angola’s Capital Markets Commission Chairman Elmer Serrão urged adaptable, proportionate oversight of financial innovation, including artificial intelligence and cloud computing. He emphasized cooperation among regulators, market operators and technology companies to enable innovation while protecting financial stability, investors and market integrity.
MAS outlines existing resilience guidelines and supervisory expectations for e-wallet providers
In a parliamentary reply, Monetary Authority of Singapore Chairman Gan Kim Yong outlined existing guidelines covering system availability and service recovery for regulated e-wallet providers, whose resilience measures are assessed through supervisory reviews that consider the impact on businesses and customers.
Central Bank of the UAE outlines 2027 regulatory priorities with focus on governance and financial sector resilience
The Central Bank of the UAE presented its 2027 regulatory priorities and 2027–2029 strategy to bank and insurance CEOs, with discussions covering suitability, governance and resilience. Cybersecurity and fraud prevention capabilities were also addressed, alongside supervisory reviews of digital assets and artificial intelligence in banking and resilience, digitalisation and climate in insurance.
Dutch Authority for the Financial Markets flags AI and cyber risks, plans stricter oversight of management notifications from 2027
The Dutch Authority for the Financial Markets (AFM) has highlighted the need for responsible AI use and stronger cyber risk controls in its latest review of financial service providers. From 2027, it will scrutinize management notifications more strictly, as individuals too often start work before their fitness and propriety have been assessed. Providers can also support consumers’ financial resilience by improving their understanding of their finances and helping them make sound financial choices.
Norway's Department of Finance requests review of proposal to strengthen contingency arrangements for bank deposit data
Norway's Department of Finance has referred a project recommendation on strengthening contingency arrangements for deposit data to the Banking Crisis Committee for assessment, with the aim of maintaining data availability during severe bank operational failures such as those caused by cyberattacks.
Bank of Central African States and National Bank of Cambodia set cooperation priorities for 2027 to 2028
The Bank of Central African States and the National Bank of Cambodia reported a bilateral cooperation agreement signed on Oct. 22, 2026. Priorities for 2027 to 2028 include capacity building, digital transformation, financial inclusion, and innovation and cybersecurity. The program also covers oversight, modernization and interconnection of payment systems and instruments.
Polish Financial Supervision Commission agrees with NASK to enhance financial sector cyberthreat monitoring
The office of the Polish Financial Supervision Commission (KNF) has agreed with the NASK National Research Institute to develop moje.cert.pl for financial sector cyberthreat monitoring. Planned enhancements cover access to threat information, review of aggregated data and detection of vulnerabilities and misconfigurations, alongside tools for threat notifications and communication.
South Korea's Financial Services Commission urges stronger wallet safeguards after USD 6.3 billion Binance hack
South Korea's Financial Services Commission reviewed domestic virtual asset providers' safeguards after Binance reported approximately USD 6.3 billion in unauthorized hot wallet withdrawals. Providers were urged to increase cold wallet holdings and reserves, continuously monitor deposits and withdrawals, and promptly share information on unusual access attempts. Checks also covered access controls, withdrawal approval and signing systems, and insurance coverage.
Thailand's Securities and Exchange Commission directs digital asset operators to assess risks from reported overseas Binance cyber incident
Thailand's Securities and Exchange Commission has directed licensed digital asset operators to check customer asset security and manage risks following a reported cybersecurity incident at the overseas Binance platform. Operators reported no significant impact on assets held in custody in Thailand, while the facts and extent of any damage remain under review. Operators must immediately report incidents that could affect assets or services and keep customers informed of service impacts or restrictions.
Dubai Financial Services Authority publishes analysis of cyber resilience risks from frontier and agentic AI
The Dubai Financial Services Authority has published a paper assessing how frontier and agentic artificial intelligence could accelerate cyber activity and complicate financial institutions’ control of critical operations. It examines timely reduction of cyber exposures, dependencies beyond direct suppliers, limits on delegated authority and reliable recovery. The analysis does not create or signal supervisory requirements.
Abu Dhabi Global Market Financial Services Regulatory Authority monitors Binance security incident involving unauthorized asset withdrawals
The Abu Dhabi Global Market Financial Services Regulatory Authority is monitoring the Binance security incident involving unauthorized asset withdrawals on Oct. 30, 2026, and engaging closely with the exchange.
Hellenic Capital Market Commission and Bank of Greece adopt revised TIBER-EU framework for cyber resilience testing
The Hellenic Capital Market Commission and the Bank of Greece have adopted the revised TIBER-EU framework nationally through TIBER-GR for threat-led penetration testing that meets DORA requirements. Testing simulates cyberattacks on critical production systems and functions and is mandatory for financial entities within DORA’s scope that meet the applicable eligibility criteria. A joint team drawn from both authorities will coordinate and support the tests.
EIOPA flags elevated market and cyber risks as insurance risks fall to a low level
EIOPA’s October 2026 Insurance Risk Dashboard finds market and cyber risks still elevated, with bond market volatility and concerns about artificial intelligence and geopolitical spillovers contributing to those assessments. Insurance risks dropped to a low level as premiums grew across life and non-life business and the upper tail of the loss ratio distribution improved. Macroeconomic risks remained medium, but their outlook deteriorated, while digitalisation and cyber risks also showed a worsening outlook.
EIOPA flags elevated market risk and worsening cyber outlook for occupational pension funds
The European Insurance and Occupational Pensions Authority’s October 2026 risk dashboard identifies elevated market risks for European occupational pension funds, with wider bond spreads, increased bond volatility and high valuations creating vulnerabilities. The outlook for digitalisation and cyber risks is worsening amid threats associated with frontier artificial intelligence models and international conflicts. Stronger financial positions among defined benefit schemes in the second quarter of 2026 lowered reserve and funding risks.
Australian Treasury outlines UK cooperation on private markets, operational resilience and AI
The Australian Treasury outlined UK–Australia agreement to continue sharing information on private market developments, supervisory experience and emerging risks. Discussions also examined how comparable third party oversight regimes could reduce duplication, alongside agreement to cooperate on AI’s operational resilience implications. Both countries will explore cooperation on AI in financial advice, including balancing access and consumer protection.
Bank of Greece and Hellenic Capital Market Commission adopt revised TIBER-EU framework for cyber resilience testing
The Bank of Greece and the Hellenic Capital Market Commission have adopted the revised TIBER-EU framework nationally for threat-led penetration testing of financial entities’ critical systems and functions. Testing is mandatory for entities within the scope of Regulation (EU) 2022/2554 that meet the applicable eligibility criteria. A team drawn from both authorities will coordinate and support tests under the TIBER-GR framework.
South Korea’s Financial Services Commission tests disaster recovery under combined sinkhole and cyberattack scenario
South Korea’s Financial Services Commission conducted an exercise testing financial service continuity during a simulated sinkhole, data center fire and cyberattack. The drill examined switching operations to a disaster recovery center in Busan and restoring securities trading and the Seoul data center. Findings will inform updates to the critical infrastructure crisis management manual and improvements to risk assessment capabilities.
Central Bank of Russia outlines new security requirements for financial sector IT outsourcing and cloud services
The Central Bank of Russia has detailed legislation approved by the Federation Council establishing information security requirements for financial institutions’ IT outsourcing and cloud providers. Supervised institutions must notify the regulator before engaging providers, and both parties will be liable for disclosure of information they exchange. Providers must confirm compliance annually through an appropriately licensed audit firm.
South Korea's Financial Services Commission investigates Bithumb breach and plans stronger safeguards for virtual asset firms
South Korea's Financial Services Commission has begun joint investigations into cyber incidents including Bithumb's suspected employee data leak, while customer exposure remains under investigation. Consumer protection measures would include full compensation if losses occur. The commission also plans security safeguards comparable with those at banks for virtual asset firms and other sectors outside the Electronic Financial Transactions Act, continuing an emphasis in its earlier cybersecurity proposals.
Bank of Japan announces November 30 seminar on security risks from AI agents
The Bank of Japan’s Institute for Monetary and Economic Studies will hold an online seminar on November 30 examining security risks from artificial intelligence (AI) agents. Topics include misinterpreted instructions, actions beyond authorized limits, inadequate verification and external attacks. The sessions will address safeguards for autonomous agents and changes to web security design and operation needed to accommodate their access.
Indonesia's Financial Services Authority outlines governance cooperation priorities with 27 professional associations
Indonesia’s Financial Services Authority (OJK) outlined cooperation priorities with 27 governance professional associations covering integrity, digital and artificial intelligence risks, and sustainability. At a governance forum, OJK Chair Friderica Widyasari Dewi requested focused working groups to develop measurable joint programs. Cooperation will include professional training, certification, stronger standards and joint policy research.
Bank Negara Malaysia revises framework for electronic trading platforms in wholesale money and foreign exchange markets
Bank Negara Malaysia has revised its framework for electronic trading platforms in wholesale money and foreign exchange markets, replacing the 2019 framework from Nov. 1, 2026. Operators require prior approval, independent assurance of controls and advance notification of changes to services, alongside restrictions on trading as principal or agent. The framework also sets requirements for user due diligence, data protection, cloud risk management and operational continuity, with immediate reporting of disruptions and surveillance of trading irregularities.
Bank of England and UK Financial Conduct Authority survey finds AI use at 86% and gaps in ongoing monitoring
The Bank of England and the Financial Conduct Authority’s fourth AI survey found that 86% of respondents use the technology, while agents with broad autonomy remain rare. External providers supplied 53% of systems assessed as highly material, but ongoing data risk monitoring and review controls were each reported by only around half of respondents. Cyber risk amplification and dependencies on critical third parties were the leading systemic concerns and were expected to intensify over the next one to three years.