Top central bankers and finance ministers from the world’s most powerful economies will simulate the collapse of a major global bank this weekend in a high-stakes war game designed to test their crisis response.
The exercise, a behind-closed-doors “desktop” stress test, will take place in Washington on Saturday at the offices of the US Federal Deposit Insurance Corporation (FDIC). It will involve the most senior officials from the US Federal Reserve, the Treasury, the Bank of England – including Governor Andrew Bailey – and their counterparts from the European Central Bank and the European Commission.
Their gathering coincides with the end of the International Monetary Fund and World Bank spring meetings, where finance ministers and regulators have been sounding the alarm over a suite of emerging threats to financial stability.
Foremost among the concerns driving the urgency of this weekend’s drill is the rapid evolution of artificial intelligence. Discussions in Washington have centred on the specific risks posed by advanced AI models like Anthropic’s “Mythos”, which experts warn could expose critical weaknesses in banks’ cyber defences with unprecedented speed.
Andrew Bailey, who also chairs the global Financial Stability Board, described the challenge as “a very serious challenge for all of us,” noting that “it reminds us how fast the AI world moves.”
The specific AI threat prompting alarm
The anxieties are not theoretical. Anthropic has stated that its Mythos model has already identified thousands of high-severity vulnerabilities in major software systems. Its ability to chain these flaws together at scale presents a novel threat, potentially increasing the frequency and impact of cyber attacks on financial institutions.
Beyond direct cyber threats, regulators fear AI could amplify risks across markets. The widespread use of common, opaque AI models by different firms could synchronise trading behaviour and amplify market correlations during a crisis. Furthermore, the “black box” nature of some AI, with its complex and unexplainable decision-making, creates significant model risk for banks that rely on it.
Access to such powerful AI is currently restricted to around 40 companies for testing, including major financial institutions like J.P. Morgan Chase. The US Treasury is seeking its own access to evaluate the model, while Anthropic is preparing to offer it to UK financial institutions, forcing regulators to scramble to develop appropriate safeguards.
Alongside AI, officials are scrutinising other systemic vulnerabilities. These include the booming but opaque $2 trillion private credit sector, where regulators are intensifying oversight and demanding greater data disclosure from banks with exposure to it. Geopolitical instability, particularly the war on Iran and its potential to disrupt global energy supplies and shipping lanes, adds another layer of complexity to the stability picture.
This confluence of new and old risks forms the backdrop for Saturday’s simulation, which is formally known as a Trilateral Principal Level Exercise (TPLE). These exercises have been a regular feature since 2014, involving US, UK and European Banking Union authorities, and are deliberately timed to leverage the rare physical gathering of top officials after the IMF meetings.
The FDIC, which acts as the insurer of US bank deposits and the receiver for failed banks, stated the exercise aims to “enhance understanding of each jurisdiction’s resolution regime for global systemically important banks (G-SIBs), strengthen coordination on cross-border resolution, and promote confidence in and commitment to the orderly resolution of G-SIBs.”
The shadow of the 2008 collapse of Lehman Brothers, which triggered the global financial crisis, looms large over the event. Lehman’s failure, with debts of $619 billion, led to a liquidity crisis, halted lending, and wiped trillions from global stock markets. Since then, regulators have built frameworks to handle “Lehman-scale problems” at the world’s most important banks.
The weekend’s wargame focuses specifically on these G-SIBs – the 29 banks deemed too big to fail without causing widespread damage. The US, UK, and EU jurisdictions are home to 18 of these institutions, including US giants like JPMorgan Chase and Citigroup. These banks are subject to stricter rules, including holding more capital and creating detailed “living wills” to plan for their own orderly failure without taxpayer bailouts.
By simulating a cross-border collapse, the central bankers and treasury principals are not just testing legal frameworks but practising the vital personal coordination that would be essential in a real-world crisis, aiming to ensure a repeat of 2008’s chaotic aftermath can be prevented.
