Bank of England floats ‘kill switch’ for agentic AI trading

TL;DR:

• Bank of England Deputy Governor Sarah Breeden has signalled the need for bespoke rules to govern agentic AI in finance, a shift from the Bank’s long-held view that existing frameworks suffice.
• Options under consideration include market-wide “kill switches” that would halt trading if faulty AI models trigger a meltdown, plus “enhanced recovery” letting one bank take over another’s core functions.
• A Cambridge survey cited by Breeden found 52% of finance firms already use agentic AI.

The Bank of England has, for the first time, openly questioned whether its rulebook can contain the risks of AI that acts on its own. Speaking at the European Central Bank forum in Portugal, Sarah Breeden said rapid progress in agentic payments and trading had exposed gaps that may demand a more sophisticated response.

A shift in tone from the Bank

“Our frameworks were not built to contemplate autonomous agents, and relying on a human in the loop for all agent actions is unlikely to be realistic,” Breeden said. That marks a notable change: the Bank had previously insisted existing supervision was adequate. Her central worry is correlated behaviour — if many agents respond identically to the same trigger, they could amplify volatility in a stress event, “especially if their objectives drift from original goals”.

The intervention lands amid a busy stretch for UK financial regulators wrestling with AI. The FCA has been rethinking regulation for the AI age and eyeing an agentic watchdog, while the Bank for International Settlements warned that the AI boom itself risks a financial crash. Breeden’s kill-switch language pushes that debate towards concrete mechanisms rather than principles.

For UK firms deploying autonomous agents in markets, the direction of travel is clear: expect circuit breakers, severity thresholds and recovery obligations to feature in future supervision. The Financial Stability Board made a similar call in June for tighter safeguards against agentic AI.

Looking forward

Breeden framed the measures as options rather than firm policy, but the Bank rarely trails ideas it has no intention of pursuing. With more than half of finance firms already running agentic systems, the question is less whether bespoke rules arrive than how prescriptive they become — and whether a “kill switch” can act fast enough to matter.