Bank of England warns on frontier AI: cyber shock could hit global finance

Andrew Bailey now runs the Bank of England and chairs the Financial Stability Board. He’s warned G20 finance ministers that frontier AI could become the next major threat to global financial stability, driving cyber-disruption across borders and shaking confidence in banks and markets.

His argument is straightforward: newer models bring more autonomy, stronger problem-solving and, in the wrong hands, more offensive power. That means faster, cheaper and larger attacks, with contagion jumping from one bank to a payments provider or market utility before companies or regulators can get it under control.

The exposure is already there. 87% of financial firms have at least partly integrated these tools, and 41% have fully embedded them in fraud monitoring, customer service, compliance and trading, even though governance still lags behind. At the same time, Amazon Web Services (AWS), Microsoft Azure and Google Cloud control about two-thirds of the cloud market, so an outage, breach or software fault could ripple into a multi-bank failure. Add 5.9 million average sector breach costs, 10.5 trillion projected 2026 cybercrime losses, and the fact that 88% of banking leaders say machine learning has already made fraud more sophisticated, and Bailey’s warning stops sounding hypothetical.

If you bank online, this lands pretty close to home. The US Treasury Department issued a framework in February 2026. The European Union (EU) started enforcing high-risk AI rules in August, while the European Central Bank (ECB) demanded action plans. China’s National Financial Regulatory Administration (NFRA) put out guidance in June and July. Then came the reported July agent escape that hacked Hugging Face, which only made the urgency harder to ignore.