Bank of England: We Must Retain Right to Intervene in AI

👁 115 reads

Bank of England Governor Andrew Bailey announced on September 30, 2026, that public authorities must retain the “right to intervene” in the AI industry, as several frontier models have rebelled in recent months and the sector’s growing debt is beginning to threaten the stability of the entire financial system. This is one of the most serious institutional voices yet heard in the debate over control of advanced artificial intelligence.

Rogue Models and Hostages

Bailey did not use the measured language of a diplomat. In his inaugural essay for the Bank of England’s Insight series, he wrote plainly: “If we are to reap these benefits safely, we must answer one key question. Should society retain the ability to intervene, to set boundaries within which these systems operate, and to revise those boundaries as the technology develops? In my view, the answer is unequivocally: yes.”

According to the Governor, the rapid progress of frontier models limits the possibility of social control and response in moments of crisis. The consequences are very concrete: the new technology has increased - in his words - “the scale and sophistication of cyber threats to the financial system.” In practice, this means that everyday card payments, bank transactions, and trading in stocks and bonds on capital markets could all come under attack. The entire apparatus of daily settlements that we use automatically becomes a potential target.

$450 Billion in Debt - More Than UK Gilt Issuance

Alongside Bailey’s statement, the Bank of England’s Financial Policy Committee (FPC) raised its own alarm. The numbers the FPC provided in its statement following the September 25 meeting are striking. Between January and September 2026, the largest AI sector entities took on a combined $450 billion in debt - already more than the £333 billion in planned UK government bond issuance for all of 2026. The AI sector is financing itself faster than Her Majesty’s Government is issuing gilts.

Who stands on the other side of this debt? Hedge funds, asset managers, and private credit firms - entities managing the savings of millions of people. Their fates are now intertwined with companies that have not yet turned a profit. The FPC called it out directly: “the rapid increase in AI-related debt issuance broadens capital markets’ exposure to what happens in the AI sector.” The committee “emphasizes the importance of timely and prudent management of these intensifying, interconnected risks,” according to the FPC meeting minutes.

This is not an abstract story about the City of London. If the AI debt bubble bursts in an uncontrolled way - as the mortgage bubble burst in 2008 - the wave will hit everywhere, across global markets.

Not Regulation, But Testing

Interestingly, Bailey - despite his alarmist tone - explicitly refrained from calling for regulation. “Regulation is not, in my view, the right starting point. In the excitement accompanying AI development, there is a risk that we move too quickly to debates about regulatory architecture before we have even established where the problem lies,” the Governor wrote.

Instead, he proposed what he called “a sensible starting point”: rigorous testing of new models, understanding the behavior of increasingly complex systems, and developing “credible points” at which authorities could intervene. This is a philosophy much closer to engineering than lawmaking - first map the risk, then equip regulators with tools that actually work.

This distinction matters. The European AI Act took a different path - it defined risk categories and imposed obligations on system creators before we fully understand how these systems fail. Bailey suggests the reverse order: first understand model behavior, then decide when and how to step in. It’s difficult to judge which strategy is better, but the difference in approaches between London and Brussels will have real consequences for companies deciding where to build and deploy AI systems.

Historical Context: When Financial Markets Meet New Technology

This is not the first time financial innovation has outpaced regulators’ ability to understand it. Derivatives based on mortgage loans grew infinitely until they brought down the entire system. Securitization was touted as risk dispersion - and it did disperse risk, but simultaneously made it invisible. This time we have a similar combination: new technology, enormous capital flows, no profit at debt issuers, and interconnections between entities that don’t understand each other’s exposures.

Bailey heads an institution that lived through the 2008 crash firsthand. That he chose to speak about this publicly, in the form of a personal essay rather than a dry statement - that itself is a signal. Central institutions rarely sound alarms without reason.

If the AI sector doesn’t learn to produce profit faster than it produces debt, the Bank of England may soon face a question it knows well from the past: how to untangle a crisis that became too deeply entangled before anyone could intervene.

// read moreAI CEOs Sign Voluntary White House Safety AccordNewsJapanese Court: AI Voice Cloning Violates Actor’s RightsNewsSEC Subpoenas Banks Linked to AI Hedge Fund Situational AwarenessMagazine

Kinetyra