The Bank of England chief has warned that advanced artificial intelligence models could create new risks for global financial stability. In particular, these risks could arise by making cyberattacks faster, larger and harder for financial institutions to contain.
Andrew Bailey, who also chairs the Financial Stability Board, raised the concern in a letter to G20 finance ministers and central bank governors. He said so-called frontier AI models are developing increasingly sophisticated autonomy, problem-solving capabilities and potential threat capabilities.
Bailey identified cyber risk as the most immediate financial stability concern associated with frontier AI. Moreover, he warned that the technology could materially change the speed, scale and economics of cyberattacks. This could potentially undermine confidence across financial markets.
The Financial Stability Board’s August 2026 letter also highlights the concentration of important financial services among third-party technology providers. If an advanced AI-enabled cyberattack were to disrupt a widely used provider, the consequences could spread across multiple institutions. This disruption could occur at the same time.
Bailey’s warning comes as financial institutions increasingly experiment with AI for trading, analysis, customer services, compliance and other operations. The technology can improve efficiency, but greater dependence on shared digital infrastructure also creates new points of failure.
Why Frontier AI Creates a New Financial Stability Risk
Frontier AI refers to the most advanced artificial intelligence models currently being developed. Their growing ability to perform complex tasks with limited human intervention is changing the nature of both opportunity and risk.
For financial institutions, the concern goes beyond whether an AI system produces an incorrect answer. The larger issue is what happens when increasingly capable models interact with critical financial infrastructure, software systems and networks.
A sophisticated AI model could potentially help defenders identify vulnerabilities more quickly. At the same time, the same capabilities could allow malicious actors to discover weaknesses, automate attacks and operate at a scale that would be difficult for human attackers to match.
The Bank of England’s Financial Stability Report has warned that recent advances in frontier AI have increased risks related to cyber and operational resilience. The central bank said advanced models are increasingly capable of identifying and exploiting software vulnerabilities across multiple stages of an attack.
That creates a race between attackers and defenders.
Financial firms would need to identify vulnerabilities, apply security patches and respond to incidents more quickly. In addition, if several institutions rely on the same cloud provider, software component, data service or AI infrastructure, a single disruption could have consequences beyond the organization initially targeted.
Bailey said financial institutions and technology providers therefore need stronger vulnerability management, response and recovery capabilities. They also need to prepare for scenarios in which several firms or shared technology dependencies are disrupted simultaneously.
The problem is particularly complicated because financial markets are highly interconnected. Banks, insurers, asset managers, exchanges and other institutions depend on networks of external providers. A failure at one critical supplier can therefore affect firms that may appear unrelated.
The Bank of England has already been examining these risks. Its financial stability work identifies several channels through which AI could affect the financial system. These channels include greater use of AI in core financial decisions, increased adoption in financial markets, dependence on external AI providers and a changing cyber threat environment.
The pace of technological development makes the problem harder to manage. Regulators may be attempting to establish safeguards while the capabilities of the underlying technology are changing rapidly.
Bailey said many jurisdictions still do not have protocols capable of managing the development, release and deployment of advanced frontier AI models. That regulatory gap could increase risks not only for financial institutions but for the wider economy.
AI Is Not the Only Risk Facing Global Markets
Bailey’s warning does not focus exclusively on artificial intelligence. His letter also points to several existing vulnerabilities that could make financial markets more sensitive to a new technological shock.
Among them are fragilities in sovereign debt markets, increased use of leverage by investors in equity markets and stretched asset valuations. This is particularly true in investments connected to AI.
That combination creates an important distinction between AI as a technology risk and AI as a market risk.
The first involves cybersecurity, operational failures and dependence on technology providers. Meanwhile, the second involves the possibility that investors have become overly exposed to companies or assets whose valuations depend heavily on expectations for future AI growth.
The Bank for International Settlements’ analysis of AI and the global economy notes that the AI investment boom is driving a large and increasingly debt-financed surge in investment. It is also affecting equity markets and economic growth.
The BIS has also warned that the productivity benefits of AI remain uncertain and uneven across countries and industries. That uncertainty matters for financial markets because expectations about future productivity can influence asset prices, investment decisions and corporate borrowing.
A sharp reassessment of those expectations could therefore create financial stress even without a major cyberattack.
AI could also accelerate market movements. Advanced systems can process information and make decisions at very high speed. Furthermore, if many market participants rely on similar models, data or signals, their responses to new information could become more closely correlated.
During normal conditions, faster decision-making can improve market efficiency. During periods of stress, however, synchronized trading can amplify price movements and make liquidity conditions more difficult.
This is why financial stability officials are increasingly looking at AI as a system-wide issue rather than simply a technology issue.
The challenge for regulators is to distinguish between useful innovation and vulnerabilities that could spread across institutions. That requires better information about how financial firms use AI. Specifically, regulators must know which external providers they depend on and how their systems would behave under extreme conditions.
Regulators Face Pressure to Prepare Before the Next Shock
Bailey’s warning also highlights a broader regulatory problem: financial authorities cannot assume that existing safeguards will automatically remain adequate as AI capabilities advance.
Financial institutions will need to strengthen cyber defenses while also testing how their systems respond to AI-driven threats. Recovery plans may need to account for incidents that develop faster and affect multiple organizations simultaneously.
The NIST AI Risk Management Framework provides one example of a broader risk-management approach for organizations developing and deploying AI systems. Although it is voluntary and not specific to financial services, the framework is designed to help organizations identify and manage risks throughout the AI lifecycle.
For financial regulators, international coordination will be particularly important.
The Financial Stability Board brings together authorities from major economies precisely because financial disruptions can cross national borders. A cyber incident involving an AI provider, cloud platform or critical piece of financial infrastructure could spread internationally much faster than regulators can respond. This could happen if they work independently.
Bailey’s warning to the G20 therefore reflects a problem that cannot easily be contained within one country’s financial system.
The United States is hosting the G20 summit in Asheville, North Carolina, this week, bringing together finance ministers, central bank governors and other senior officials to discuss global economic priorities. Frontier AI is emerging as one of the issues that financial authorities will need to consider. They will address it alongside more traditional sources of market instability.
The immediate concern is not that AI will inevitably cause a financial crisis. Rather, advanced models are changing the speed and scale at which existing vulnerabilities can be exploited.
For Bailey, that makes preparation urgent. Financial institutions and regulators need to improve resilience before a major AI-enabled disruption tests whether the safeguards currently in place are sufficient.
