Senior officials from the Bank of England and the International Monetary Fund issued separate but aligned warnings this week about the rapid advance of artificial intelligence, highlighting both the operational risks of autonomous AI agents and the systemic dangers of AI-related borrowing.
Sarah Breeden, the Bank of England’s deputy governor for financial stability, used a speech at the European Central Bank’s annual forum in Sintra, Portugal, to caution that agentic AI — systems capable of planning and executing decisions without human oversight — is evolving far faster than regulators anticipated. She noted that in 2019 the length of software tasks that leading AI models could complete was doubling every seven months; by 2024, that doubling time had shrunk to just four months. Breakthroughs in cyber vulnerability detection this spring, she said, suggest the pace may have accelerated even further. “We were surprised this Spring, and we should be prepared for further technology surprises,” Breeden said.
Breeden outlined three stages of AI progression: generative models that produce content when prompted, reasoning models that handle multi-step tasks, and now fully agentic systems that can trade securities, process payments, and respond to cyber threats with limited or no human involvement. Such systems, she warned, could lead to a financial landscape that “operates more autonomously, at scale and speed.” To counter potential flash crashes or runaway trading loops, she proposed embedding mandatory “kill switches” into autonomous trading tools — a departure from current regulations.
Cybersecurity risk dominated her remarks. Citing findings from the UK government’s AI Security Institute, Breeden stressed that the same tools that help defenders patch vulnerabilities also give attackers the ability to discover and exploit them, thereby “materially increases the chance of attacks that could harm financial stability.”
A second front of concern is the financing of AI infrastructure. The Bank’s Financial Policy Committee has observed that large tech companies are increasingly turning to debt — often in novel, complex forms — to fund data centers, chips, and computing capacity. A sharp decline in AI-related asset prices could therefore send ripples through credit markets. “The financial stability consequences of any fall in AI-related asset prices could well increase,” Breeden said, with a deeper committee assessment due on July 7.
Echoing these debt concerns, IMF Financial Counsellor Tobias Adrian warned that AI-related borrowing now poses a bigger financial stability risk than elevated tech stock valuations. While current earnings do not yet signal an AI asset bubble, the rapid accumulation of debt to fund AI investments — by governments and private firms alike — could amplify losses if sentiment shifts. Adrian’s remarks underscore the global dimension of the challenge, with the IMF flagging potential spillovers far beyond the tech sector.
Both institutions called for central banks not only to tighten oversight of AI-driven threats but also to harness AI internally to monitor the increasingly autonomous financial system.
