A senior Bank of England official is urging regulators to update oversight frameworks to address the rapid emergence of agentic AI in payments, trading, and financial operations.
Speaking at a central banking conference in Portugal, deputy governor Sarah Breeden warned that existing supervisory models were never designed for autonomous systems and may no longer be sufficient as AI agents take on more decision‑making roles.
Breeden said relying on a human in the loop for every agent action is increasingly unrealistic, especially as autonomous trading and payment systems scale. She also highlighted the growing likelihood of more severe cyberattacks, noting that policy frameworks, internal skills, and institutional structures must be prepared for “more frequent technology surprises.”
Her proposals include expanding scenario analysis, using AI to strengthen monitoring and risk management, and developing digital twins to simulate interactions between banks. She also said the BoE is evaluating “enhanced recovery” mechanisms that would allow one bank to temporarily take over another’s core processes during a disruption.
Breeden referenced potential kill switches and market‑wide circuit breakers to halt trading if faulty AI models trigger instability or meltdown. The speech follows a recent call from the Financial Stability Board for stronger safeguards around the use of AI agents in financial systems.
