G7 central bank chiefs convene Lehman-style war game at FDIC to stress-test global financial system against AI-driven systemic bank failure scenarios

Senior central bank officials from the US, UK, and EU are conducting a joint war game exercise at the offices of the FDIC (Federal Deposit Insurance Corporation — the US agency that guarantees bank deposits) in Washington, following the conclusion of IMF meetings. The exercise is designed to assess how regulators would respond to a Lehman Brothers-style collapse of a significant financial institution, with the scenario specifically incorporating AI-driven threat vectors. Bank of England Governor Andrew Bailey, who also chairs the Financial Stability Board (FSB) — the international body that coordinates financial regulation across G20 jurisdictions — has explicitly linked the war game to concerns about frontier AI models' capacity to expose flaws in bank IT systems. Bailey described the threat as 'a very serious challenge for all of us', noting that it 'reminds us how fast the AI world moves'. The exercise reflects escalating institutional concern about whether existing resolution frameworks — the legal and regulatory toolkit for managing a failing bank without triggering wider contagion — are adequate to address failure scenarios where AI-exploited vulnerabilities could cause simultaneous, correlated disruption across multiple institutions. The Bank Recovery and Resolution Directive (BRRD) in the EU and equivalent UK frameworks were designed for traditional credit and liquidity failures, not cyber or AI-catalysed systemic events. The FSB's involvement signals that the output of this exercise is likely to feed into international regulatory standard-setting, potentially resulting in new guidance or rule changes affecting how banks must document and stress-test AI-related operational risks. For UK-regulated firms, this sits squarely within the PRA's (Prudential Regulation Authority's) operational resilience regime and the FCA's (Financial Conduct Authority's) existing AI governance expectations.

Sign up to read →