Current thematic dossiers

Select a theme to view its dossier.

Artificial intelligence

This deep dive examines the latest developments in artificial intelligence, including emerging adoption patterns, key risk watch points, evolving policy and supervisory guidance, and research on the economic impact of AI.

Overview

Since the start of the year, authorities' attention has concentrated on four key AI risk watch points:

  • Frontier AI capabilities that could accelerate cyber attacks and operational disruption across shared financial sector infrastructure
  • AI-enabled fraud and impersonation that can make scams more persuasive, personalized and scalable
  • Growing concentration and opacity in third-party AI, cloud and software supply chains, increasingly treated as a strategic dependency where leading providers sit outside the home jurisdiction
  • AI-driven market, financial stability and infrastructure financing risks, including synchronized trading or redemption behaviour, stretched AI-sector valuations, opaque debt channels linked to data centers, chips and related infrastructure, and circular investment relationships among AI firms

Across these areas, the common concern is that AI may increase the speed, scale, interconnectedness and opacity of risk transmission, making threats harder for firms and supervisors to detect, attribute, contain and recover from before they affect consumers, markets or critical financial infrastructure.

What's new

The European Central Bank has published an analysis highlighting how US hyperscalers’ growing euro bond issuance increases concentration and exposure to AI shocks. These companies account for close to 8% of euro investment grade corporate bonds maturing after 10 years, while their growing presence across equity and bond benchmarks could reduce diversification. Evidence of displacement of other borrowers remains limited, but rising funding needs could constrain their access to capital and push up long-term sovereign yields.

Reserve Bank of Australia Governor Michele Bullock assessed inflation risks as tilted upward during a panel at the Australian Securities and Investments Commission Annual Forum. The bank's five interest rate increases totaling 125 basis points since the start of 2026 aim to curb excess demand rather than offset higher oil prices. Persistent oil price pressures, AI spending and signs of higher inflation expectations remain concerns, while housing weakness could offset some upward pressure.

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.

Deep dive

Frontier AI-enabled cyber and operational disruption

Authorities warn that advanced or frontier AI can sharply increase the speed, scale, accessibility and sophistication of vulnerability discovery and attack execution, allowing weaknesses in financial institutions, market infrastructures and the shared digital infrastructure underpinning payments, trading, communications and data exchange to be exploited before defenders can respond and potentially turning isolated breaches 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. 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. Anthropic initially limited access to about 50 Project Glasswing partners focused on patching identified vulnerabilities and sharing non-sensitive findings. 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, while 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.

As part of their warnings, authorities reiterated their expectations of firms, stressing that boards and senior management should treat frontier AI as a management and operational resilience issue rather than a purely technical one. Firms are moreover expected to 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.

Formal risk assessments and required action followed. The European Systemic Risk Board raised its assessment of systemic cyber risk to severe in June, from elevated in March, and issued a warning in July that frontier models may give attackers an advantage in speed, scale and sophistication in the short to medium term. The European Central Bank's Banking Supervision wrote to all significant banks requiring action plans by 31 October covering exposed ICT assets, accelerated patching, AI-enabled defences and third-party risk management, to be followed by a horizontal analysis. Three European Supervisory Authorities issued a joint statement calling for risk-based controls under DORA and began targeted engagement with critical ICT third-party providers, whose frontier AI risks will feature in 2027 oversight examinations. Comparable letters and recommendations were issued by supervisors in Guernsey, Poland, the Philippines, Denmark and New Zealand, while Sweden’s Finansinspektionen opened on-site reviews of firms’ defences, and the Bank of Italy asked supervised intermediaries for frontier AI risk plans by the end of 2026.

Evidence on the pace of exploitation has firmed up. The Bank for International Settlements reports that the cost of a full attack chain has fallen from USD 5,000–10,000 using Mythos to USD 50–100 on cheaper models and points to a jump in security fixes and critical vulnerability disclosures since April. The Monetary Authority of Singapore notes that high-severity vulnerabilities disclosed in 2026 rose sixfold to 2,200 against the preceding three-year average, although successful breaches have not risen at the same rate, and the IMF’s October Global Financial Stability Report finds that in more than 70% of cases this year software providers had almost no time to patch before exploitation began, 20 percentage points more than a year earlier. At the global level, the FSB chair has called on jurisdictions to support safe and responsible releases of frontier models and on institutions to secure the resilience of their critical technology providers.

AI-enabled fraud and impersonation

Authorities warn that generative AI can make fraud and scams more persuasive, scalable and difficult to detect by producing deepfake images, videos and voices, synthetic identities and conversational bots that impersonate trusted people or institutions, simulate credible communities and tailor solicitations to individual victims across public and private digital channels.

Authorities increasingly describe AI as an accelerant of fraud, allowing scammers to manufacture credible identities and endorsements, personalize approaches and maintain convincing interactions at greater speed and scale. Their warnings also show schemes moving beyond public advertisements and websites into closed messaging groups and private conversations, where AI-generated content and simulated social activity can build trust while making perpetrators harder to identify and investigate.

The Austrian Financial Market Authority identified deepfake celebrity videos, WhatsApp investment groups and AI-powered chatbots as emerging investment-fraud techniques in 2025. It described groups that promote trading tips and exclusive opportunities, use chatbots to simulate an active investment community, and then approach victims privately to direct them to fraudulent platforms or apps and overseas accounts or wallets. The authority received 843 fraud reports involving a record 19.6 million euros in losses, up from about 15.5 million euros the previous year, and issued 97 warnings about unauthorized providers, most linked to trading platforms. Separate warnings from the Bank of Italy and the Central Bank of San Marino illustrated the growing use of deepfakes to impersonate senior central bank officials: fabricated images, videos and articles presented the officials in television or other media settings and associated them with investment platforms or solicitations, prompting both institutions to file complaints with the relevant authorities.

Canada’s Financial Industry Forum on Artificial Intelligence Phase II report broadened the concern to voice cloning, synthetic identities and personalized scams, noting that effective deepfakes can be produced from limited information obtained through social media and that call centers and help desks can be targeted through convincing impersonation of customers or employees. The report cited an industry survey in which 91% of financial institutions said they were reconsidering voice-verification systems because of AI voice cloning, and said personalized fraud using consumer data and behavioral patterns is becoming harder for consumers to distinguish from legitimate communications. It also cited an estimate that 90% to 95% of fraud in Canada goes unreported. In May, the U.S. House Financial Services Committee said AI could be weaponized to conduct scams at unprecedented scale and speed and advanced legislation addressing scams targeting older adults, federal assessment of AI-related fraud risks and access to advanced fraud-detection tools for smaller banks and credit unions.

More recent data put numbers on the scale. The Australian Securities and Investments Commission removed more than 19,400 online scams in the year to June 2026, up 182%, including 7,051 fake investment platforms, and warns that generative AI allows criminals to build whole networks of deepfake endorsements, fabricated news, fake reviews and spoof websites, so that a basic online search is no longer sufficient to verify an opportunity. INTERPOL attributes 55% of reported cybercrime in Africa to AI-enabled methods and reports that losses have more than doubled since 2024 to USD 484 million, with synthetic identities used to bypass biometric checks to open accounts, obtain loans and register SIM cards. Responses are forming at several levels.

Third-party AI, cloud and supply-chain concentration

Authorities and industry forums warn that financial institutions’ growing reliance on a small number of AI model providers, cloud platforms and software vendors, together with opaque multi-tier supply chains, can create shared points of failure while limiting institutions’ visibility into and control over the models, data, infrastructure and subcontractors supporting critical services.

The concentration concern is rooted in the current market structure of providers. The South African Reserve Bank’s Prudential Authority observes that only a handful of companies have the technology and resources to train and run AI models, leaving financial institutions reliant on outsourcing and exposed to single points of failure. A review published by the Australian Prudential Regulation Authority in April noted heightened concentration risk at firm level, with some banks, insurers and superannuation trustees heavily dependent on a single provider for multiple AI use cases and few having demonstrated robust contingency planning or tested exit and substitution strategies for critical AI providers. European authorities have also raised the geopolitical dimension. The European Systemic Risk Board points to strategic dependency and geopolitical risks because leading AI providers are concentrated outside the EU, and European Central Bank President Christine Lagarde has warned that geopolitical restrictions could disrupt access to frontier models on which European financial institutions may increasingly depend. In their role as oversight bodies of critical ICT third-party providers under DORA, the three European Supervisory Authorities have begun targeted engagement with these providers on how they manage frontier AI risks, and expect AI-related risks to feature in their oversight examinations in 2027.

Authorities also point to limited visibility into what sits behind the AI services firms use. The Australian Prudential Regulation Authority found that, as AI capabilities are increasingly embedded within software, platforms and developer tools, upstream dependencies such as foundation models, training data sources and fourth-party service providers can be opaque, limiting institutions’ ability to independently assess model performance, bias, resilience and security. It also found limited evidence of contract provisions addressing audit rights, model updates, incident notification or changes to data handling. The Canadian Financial Industry Forum on Artificial Intelligence found visibility into fourth- and fifth-party relationships especially limited for AI services, with institutions often lacking clarity on how models were trained, what data was used and which other entities are embedded in the supply chain. It also noted that many critical providers operate outside the financial regulatory perimeter and that many AI services and associated data reside abroad.

AI-driven market, financial-stability and infrastructure financing risk

Authorities, researchers and lawmakers warn that wider use of AI in trading, investment and treasury decisions could produce synchronized or unpredictable market behavior and accelerate redemptions or funding outflows, while stretched technology valuations and opaque debt financing for data centers, chips and related infrastructure could transmit an AI-sector downturn through banks, private credit funds, insurers, pensions and capital markets.

Research and official statements have sharpened two distinct financial stability concerns: how autonomous AI systems may behave during periods of stress and how the expansion of AI infrastructure is being valued and financed.

In May, researchers writing in a European Central Bank publication used a simulated mutual fund redemption problem to show that different AI architectures could produce materially different forms of instability. Q-learning agents converged on excessive redemptions under default risk even when economic fundamentals were strong, while large language model agents were less prone to such runs but formed differing beliefs about other investors, making outcomes less predictable. The authors said the same dynamics could apply to bank runs, currency attacks and stablecoin runs. Opening the ESRB’s annual conference in October, the ECB President identified autonomous agents acting contrary to human intentions in trading as one of three AI threats warranting macroprudential monitoring, warning that reliance on a small number of models could cause firms to respond similarly to shocks, amplify price movements and make effective human oversight harder, at a time when generative AI is used by nearly nine out of ten significant euro area banks although agentic AI remains limited.

On valuations, authorities warn that the rise in AI-related share prices has left markets exposed to a disorderly correction. The FSB chair warned G20 finance ministers and central bank governors in August that rising equity market leverage, combined with stretched valuations, market concentration and AI-related optimism, could amplify a correction. The IMF’s October Global Financial Stability Report names a correction in AI-related assets as one of three plausible near-term triggers for market turmoil. Its global stress test finds banks broadly resilient, but banks holding 22% of global banking assets would fall below a 7% CET1 ratio under a scenario combining an energy shock, higher inflation and bond yields, and an AI-led correction. The ECB finds that earnings associated with AI investment have been the main force supporting US equity prices, with concentration in the S&P 500 above its historical 95th percentile, so that underperformance by a few large AI-related companies could affect the broader market.

On financing, the concern is that the build-out increasingly relies on debt and on interconnected arrangements that can obscure leverage and underlying demand. In January, Democratic members of the US Senate Banking Committee asked the Financial Stability Oversight Council to investigate more than USD 1 trillion in projected debt financing for AI infrastructure, citing the growing use of private credit, securitisations and off-balance-sheet structures. A Bank for International Settlement analysis published in October finds that other AI firms account for 55% of investment into AI firms, with almost half of these deals by value involving firms that also have a supply chain relationship, arrangements that can obscure underlying demand and expose firms to simultaneous investment losses and falling sales. An IMF technical note published the same month projects AI-related capital expenditure above USD 4 trillion by 2029, nearly three quarters of it by hyperscalers, with funding shifting from internal cash flow towards bonds, private credit and securitisation. It judges current risks manageable because the expansion remains anchored by profitable incumbents with strong balance sheets, but notes that data centre and GPU cloud providers are more constrained by debt and weak profitability, and documents 14 circular financing deals in which about USD 300 billion of equity supported USD 1.4 trillion of future revenue commitments. The note suggests that measures to improve transparency, monitor interconnected exposures and identify leverage-related vulnerabilities may become increasingly important as AI-related financing expands.

Key sources