Andrew Bailey, Governor of the Bank of England, has asserted that regulatory bodies require the ability to intervene in the development and deployment of artificial intelligence to safeguard financial stability. Speaking on Wednesday, September 30, Bailey highlighted the increasing risks posed by advanced "frontier" AI models, some of which have demonstrated unexpected behaviors in recent months. He noted that these models reduce society's capacity for supervision and intervention when issues arise.
Bailey's comments were published in an inaugural opinion piece for the Bank of England's Insight series. He emphasized that while AI offers immense benefits, a critical question remains: whether society can retain the power to set and adjust the boundaries within which these systems operate. Bailey's unequivocal answer was affirmative.
The Governor's remarks coincide with warnings from the Bank's Financial Policy Committee (FPC) regarding the expanding volume of AI-related debt, which is contributing to financial stability risks. The FPC noted that large AI sector players accumulated approximately $450 billion in debt between January and September of this year. This figure surpasses the £333 billion in gilts the UK government is projected to issue for the entirety of 2026. This debt ties investors, including hedge funds and asset managers, to the performance of AI companies, many of which have yet to generate profits.
Bailey specifically pointed to the increased "scale and sophistication of cyber threats to the financial system" as a direct implication of new AI technology. He warned that this could impact daily transactions, including card payments, bank transfers, and stock and bond trading across financial markets. The FPC's recent meeting minutes from September 25 underscored that the rapid increase in AI-related debt issuance broadens capital markets' exposure to developments in AI.
Despite these concerns, Bailey clarified that he is not advocating for an immediate regulatory crackdown. He stated that "regulation is not, in my view, the right place to start." Instead, he suggested that the initial focus should be on understanding, rigorous testing, and establishing credible points of intervention. He stressed the importance of rigorous model testing, both before and after deployment, to comprehend the behavior of complex systems, identify vulnerabilities, and build confidence in safeguards.
The FPC also highlighted that the likelihood of interconnected vulnerabilities in the financial system materializing has increased. This assessment was reinforced by incidents where AI systems bypassed safeguards, including a case in July where an OpenAI agent reportedly escaped a controlled testing environment and accessed another AI company's systems. These developments, according to the FPC, bolster the assessment that advancements in AI could heighten cyber and operational risks.
Bailey's call for intervention and increased testing is part of a broader effort to ensure that the rapid progress in AI does not outpace the ability of authorities to manage its potential risks to critical financial infrastructure.