Bank of England flags AI as a financial stability risk

TL;DR:

• The Bank of England’s half-yearly assessment names AI a growing threat to financial stability.
• It warns that investors borrowing to bet on AI, plus concentrated positions, could amplify any fall in share prices.
• The BoE judged Britain’s banking system resilient but signalled that bespoke AI regulation may be needed.

The Bank of England has put artificial intelligence at the centre of its latest financial stability warning. In its half-yearly Financial Stability Report, published on Tuesday, the central bank said investors are betting heavily that AI will pay off while the technology simultaneously widens banks’ exposure to cyberattacks, a combination it treats as a fresh systemic concern.

Exuberance meets leverage

The BoE’s worry is less about AI itself than about how the market is financing it. Alongside familiar risks, from stretched valuations to high public debt, it flagged investors, including hedge funds, borrowing to buy shares, and AI-related companies taking on heavy debt to fund their build-outs. A “reassessment of these prospects”, the bank warned, could trigger a fall in equity prices “amplified by high concentration, correlated momentum-driven positions and increased leverage”. It also cautioned that opacity around how AI firms have borrowed could deepen any crisis. For the bet to pay off, the BoE said, AI would need widespread profitable adoption, an effective infrastructure rollout and easy access to finance, the very buildout now facing grid and supply constraints.

Looking forward

The report continues a hardening of tone from Threadneedle Street. In late June, Deputy Governor Sarah Breeden signalled the need for bespoke rules to contain agentic AI, noting existing frameworks “were not built to contemplate autonomous agents”. It also lands alongside the FCA’s own landmark review of AI in financial services. Britain’s banks remain well-capitalised, and the BoE even proposed easing post-crisis capital rules to sustain lending. But with regulators worldwide now treating AI as a stability issue rather than a productivity story, the question of how to supervise it is no longer hypothetical.