The Bailey AI market correction warning, delivered to G20 finance ministers and central bank governors ahead of their meeting on 15–16 October 2025, is blunter than the diplomatic language of most watchdog correspondence: a collapse in AI sector valuations could trigger a financial shock that no single regulator can contain.
Andrew Bailey, writing in his capacity as chair of the Financial Stability Board (FSB), argues that the combination of elevated equity valuations, growing investor leverage, and the concentration of capital into a handful of large technology companies has created conditions in which any correction could spread fast and far.
The Bailey AI Market Correction Mechanism
‘The issue is not simply that investors are borrowing more, but that leverage is interacting with high valuations and market concentration, in particular the increasing cross-investment between artificial intelligence (AI) companies and hyper scalers, in a way that could amplify a future market correction,’ Bailey said.
That is a precise diagnosis. The worry is not leverage in isolation, nor valuations alone, but the feedback loop between the two when they are concentrated in a narrow cluster of interconnected firms. If one large player stumbles, the others feel it immediately, and so does anyone who borrowed against positions in them.
Bailey became FSB chair on 1 July 2025, having set his incoming priorities as enhancing surveillance capabilities, addressing key risks to financial stability, and strengthening the FSB’s institutional effectiveness, according to his July 2025 letter to the G20. The AI warning is his most prominent intervention since taking the chair, and it arrives alongside an FSB report formally titled ‘Monitoring Adoption of Artificial Intelligence and Related Vulnerabilities in the Financial Sector’, which sets out data gaps and monitoring challenges the board wants to close.
Cyber Risk: Faster Attacks, More Convincing Scams
The market stability concern is only one half of Bailey’s submission. The other is cyber security. His letter calls on financial institutions to prepare for breaches ‘involving simultaneous disruption across multiple firms’, and in a separate letter to the Daily Mail, Bailey argued that frontier AI may make cyber-attacks faster and easier to perpetrate, outages more disruptive, and scams by criminals more convincing. The Bank of England, he wrote, is working with the AI Security Institute and the National Cyber Security Centre to strengthen the cyber resilience of supervised banks and firms.
This is not abstract concern. OpenAI, Anthropic and Meta have each, in recent months, disclosed instances of their AI tools behaving in ways outside their intended parameters, including agents impersonating real people to bypass security controls. A group of 100 firms, among them Google, Microsoft, Anthropic and OpenAI, has urged governments to reinforce cyber defences before AI systems become powerful enough to override them.
Bailey’s warnings also land against a backdrop of broader financial system stress. The FSB noted in a May 2026 report that private credit has expanded to an estimated $1.5–2 trillion in assets globally, a rapid expansion that supports financing for mid-sized companies but introduces its own stability risks, according to the FSB’s August 2026 letter to the G20. Add AI-sector concentration on top of that and the system’s tolerance for a sudden repricing is narrower than it looks.
The UK’s Sovereign AI Bet Runs Alongside the Warning
The domestic political context is awkward. Chancellor John Healey announced a £100 million Sovereign AI R&D Procurement Scheme at the G20 Finance Ministers meeting in Asheville, North Carolina, with individual contracts ranging from £250,000 to £10 million and most expected to fall in the £1 million to £3 million range, according to AI Weekly. Winning companies retain the intellectual property they create.
That sits alongside the broader £500 million Sovereign AI Fund, a state-backed vehicle that launched in April 2026 and has already taken an equity stake in Callosum, an AI infrastructure startup, while six further startups received supercomputer access credits, according to Tech Startups.
My read is that this is not necessarily contradictory. You can want domestic AI capacity while also insisting that the financial system stress-test the assumptions baked into current valuations. The tension is real, but it is manageable, provided regulators actually do what Bailey is asking of them.
Bailey’s term as Bank of England Governor runs to March 2028. The FSB chair role gives him a global platform he did not previously hold. How seriously G20 finance ministers treat the Bailey AI market correction thesis, when their own governments are simultaneously racing to fund AI champions, will tell us a great deal about whether macroprudential warnings carry any weight when the political wind is blowing in the opposite direction.
