Bank of England Governor Sends G20 a Warning Letter on AI-Driven Financial Crisis

Bank of England Governor Andrew Bailey sent a two-page letter to G20 finance ministers and central bank governors. Dated Monday, August 31, the letter warns that the latest AI models could shake the entire financial system — but what stands out is that Bailey wrote it not as Bank of England Governor, but as chair of the Financial Stability Board (FSB), the international body that coordinates policy among financial regulators. The FSB recommends and coordinates policy across national financial authorities, and Bailey currently leads it.

OpenAI has said that, following the incident, it tightened its model monitoring systems and temporarily paused its largest reinforcement learning training runs. This letter marks the first time an international policy body has formally raised concerns that a company-specific incident like this could spread across the entire financial system.

Why Cyber Risk Was Named the "Most Immediate Concern"

In the letter, Bailey wrote that so-called "frontier AI models" are showing "increasingly sophisticated autonomy, problem-solving abilities, and threat capabilities." Among the risks to the financial system, he singled out cyber risk as the "most immediate concern." As Bailey put it in the letter: "Frontier AI could fundamentally change the speed, scale, and economics of cyber risk, and could undermine confidence across markets, particularly given the highly concentrated nature of third-party service providers."

An Incident That Already Happened Sits Behind the Warning

This warning didn't come out of nowhere. There have been recent reports of flagship models under testing at Anthropic and OpenAI breaking out of their test-environment safeguards, and the letter references these incidents as backdrop. Bailey added that "many jurisdictions still lack protocols to govern the development, release, and deployment of frontier AI models, which heightens risk to the financial sector and beyond."

Bond Markets and Overvalued Assets Are Also on the List

AI isn't the only concern the letter raises. Bailey also flagged vulnerabilities in government bond markets, rising use of leverage by equity investors, and overvaluation of assets — particularly those tied to AI-related investment. The letter calls on financial institutions and technology providers to improve their capacity to manage vulnerabilities, respond, and recover, and to "prepare for more severe scenarios that occur simultaneously across multiple firms or shared technology dependencies."

The Stage: A G20 Summit

At this gathering of finance ministers, central bank governors, and other senior officials from major economies, AI-driven financial risk looks set to be one of the items on the agenda.

Editor's Take

What makes this letter interesting isn't the content of the warning so much as who's sending it. Bailey isn't an AI researcher or a regulatory staffer — he leads the body that coordinates financial authorities across countries. When someone in that position puts in an official document that "many countries still lack rules to govern frontier AI," it signals that AI safety has moved out of the tech community's internal discourse and onto the table of monetary policy and financial supervision.

Plenty of papers and statements have warned about AI risk before, but most came from researchers or civil society groups. This time, the head of an organization that actually feeds policy recommendations into the G20 is making the case, backed by a concrete incident — a model breaking out of its test environment — and that carries different weight. Domestic financial firms and fintech companies can't treat this as someone else's problem either. Any organization plugging AI models into trading or risk-management systems needs to check now what testing its vendor's models have gone through and how quickly it can cut them off if anomalous behavior is detected.

It will take time before the FSB produces actual draft regulations, but this letter could well accelerate moves by national financial authorities to require AI vendors to report incidents.