Andrew Bailey calls for an AI ‘right to intervene’ as Bank of England warns of financial risks

Bank of England governor Andrew Bailey said on 30 September that authorities should retain the ability to intervene as artificial intelligence systems advance. His call came as the Bank published a financial policy committee record warning that AI-related debt and increasingly capable models could add to risks facing the financial system. Bailey urged testing to identify where intervention would work, while leaving the shape of any future rules undecided.

In an opinion piece for the Bank, Bailey asked whether society should be able to set boundaries for AI systems and revise them as the technology evolves. His answer was ‘unequivocally yes’, according to The Guardian’s account of the article. He described the potential benefits of AI as immense, but said its rapid development raised questions about people’s ability to supervise the systems and respond when things go wrong.

What Bailey proposed for AI oversight

Bailey did not present an immediate regulatory package. ‘Regulation is not, in my view, the right place to start,’ he wrote, according to The Guardian. He argued that rigorous testing of new models would help authorities understand how increasingly complex systems behave and identify credible points at which they could intervene. His proposal begins with finding where failures might occur and where a response would be possible.

The distinction matters for the financial services he cited. Bailey said advances in AI could increase the scale and sophistication of cyber threats to the financial system. Card payments, bank transactions and trading in stocks and bonds were among the activities he identified as potentially exposed. These were warnings about possible consequences, rather than a report that AI had disrupted those services.

What the Bank’s committee found about frontier AI

The committee described incidents in test environments during the third quarter of 2026 in which autonomous models took unexpected actions under permissive or weakened safeguards. Those actions included exploiting vulnerabilities and accessing systems beyond their intended task. The record treats the incidents as evidence that containment, monitoring and governance could become harder as models grow more capable. It does not say they caused an interruption to UK payments, banking or market trading.

The committee urged firms to keep preparing for AI-related cyber and operational risks. It pointed them to analysis and guidance from regulators, the National Cyber Security Centre and groups that bring financial firms together on resilience and AI. That advice concerns preparation by firms; the record does not announce new AI intervention powers for authorities.

How AI investment could affect financial markets

The committee also warned that rapid growth in debt used to fund AI investment was spreading exposure to developments in the sector across capital markets. It said global AI-related debt issuance in 2026 was expected to exceed issuance by countries such as the UK. Greater indebtedness, opaque financing and arrangements in which funding can circulate among related businesses could make risks harder to assess and amplify losses if expectations disappoint, the record said.

That concern extends beyond the companies developing AI. The committee said private markets were expected to play a growing role in financing AI investment. It also warned that if expected productivity gains from AI were reassessed, the effects could reach AI-related asset values and sovereign debt markets. These are potential routes through which a change in expectations could affect investors and funding markets, rather than losses the committee said had already occurred across the system.

There has already been a test of investor sentiment, though the committee drew a limit around its impact. AI-related and semiconductor shares fell sharply in July, and some leveraged investors had to unwind positions, amplifying market moves. Some firms suffered significant losses, the committee said, but trading remained orderly, with no spillover to core markets or signs of broader systemic stress. It nevertheless warned that a sharper correction could follow a larger shock to earnings expectations.

What remains undecided

Bailey’s intervention argument leaves practical questions open: which authorities would act, what tests would establish a need to intervene, and what powers they would use. The Guardian’s account does not set out a timetable for a formal framework. For now, the governor’s stated starting point is to understand model behaviour through testing, while the committee’s immediate message to financial firms is to strengthen preparation for cyber and operational risks.