Bank of England

The Bank of England built its AI-displacement worst case on 500,000 extra unemployed against a vacancy base above 730,000. UK vacancies fell to 707,000 by the ONS's 18 June bulletin, crossing that threshold, with payrolled employment down 138,000 year on year .

The breach exposes a gap in the Bank's own evidence base: the ONS still publishes no AI-attribution layer, so the mechanism the Bank's model assumes has technically arrived without anyone able to confirm it is the cause, leaving the worst case in-territory but unverified.

The Bank of England's Financial Policy Committee directed the Bank and the Financial Conduct Authority, in its April 2026 record, to do further work on agentic AI, systems executing multi-step financial decisions without human confirmation, in payments and markets, judging the systemic risk 'likely to increase rapidly' given that 75% of UK financial firms already deploy AI .

This is the Bank's operational-stability mandate, distinct from its market-correction warning: a correlated failure spreading through payments infrastructure that 75% of firms already run on AI could outpace any existing circuit-breaker, pushing the Bank toward mandatory rather than voluntary supervision.

The Bank of England warned on 22 March 2026 that AI technology-firm overvaluation risked a global market correction, noting the five largest US tech firms had committed $650-690 billion to AI infrastructure in 2026 while Meta's free cash flow was forecast to fall by up to 90% .

For the Bank, the exposure sits closer to home than the tech sector itself: UK pension funds hold roughly £1.5 trillion with significant weighting in AI-adjacent US stocks, and its Financial Policy Committee can force UK institutions to hold additional capital against that concentration should the correction materialise.