Bailey Urges Vigilance as AI Debt Reaches Record Levels

Andrew Bailey, Governor of the Bank of England, has emphasized the need for society to retain the ability to intervene and set limits on frontier AI systems. He argues that a regulatory clampdown is not the solution, but rather rigorous testing of models before and after release. This would allow authorities to identify 'credible points of intervention' in case something goes wrong.

Bailey's concern is recursive learning, where models refine themselves on their own outputs and risk becoming a closed loop that governs itself. He cites the AI Security Institute as making good progress but urges them to quicken the pace. Bailey also highlights the link between AI and cyber risk, stating that payment networks, market infrastructure, and banks can no longer be assessed separately from AI.

The Bank's Financial Policy Committee (FPC) has published a warning about AI-related borrowing, which it says is spreading exposure across capital markets. The FPC cites a Morgan Stanley estimate that worldwide AI-related debt issuance reached £339bn ($450bn) by early September, more than twice the 2025 total. In sterling markets, AI hyperscalers accounted for 47% of corporate bond issuance so far in 2026.