Bank of England Deputy Governor Sarah Breeden has warned that increasingly autonomous artificial intelligence systems could amplify cyber disruption, market volatility and payment failures, placing the technology at the centre of financial-stability discussions at the European Central Bank’s Forum on Central Banking in Sintra. The panel was chaired by ECB Executive Board member Isabel Schnabel and included International Monetary Fund Financial Counsellor Tobias Adrian, Apollo Global Management chief economist Torsten Slok and University of Pennsylvania professor Itay Goldstein.
Breeden said the shift from generative tools towards agentic systems capable of chaining actions could make finance operate more autonomously, at greater scale and speed. She identified cyber attacks, automated trading and AI-led payments as the closest areas of concern, while also recognising AI’s potential to support long-term growth, improve risk monitoring and strengthen defensive capabilities.
The funding behind AI is now part of the stability debate. The Bank of England’s Financial Policy Committee concluded in April that infrastructure investment had largely been financed through large technology companies’ cash flows and equity, but debt funding was rising rapidly and becoming more complex. The Bank for International Settlements has also warned that an abrupt end to the AI investment boom could expose fixed-income markets, particularly where hyperscalers, AI laboratories and data-centre contractors have issued debt or entered opaque financing arrangements.
Cyber resilience presents the most immediate operational issue. In May, the Bank of England, Financial Conduct Authority and HM Treasury set out measures intended to help financial institutions prepare for frontier AI-related cyber threats. Breeden argued that authorities should plan for simultaneous disruption across several firms, strengthen system-wide stress testing and improve coordination with industry. The IMF has reached a similar conclusion, warning that shared software, cloud providers and AI models can allow one exploited weakness to affect several institutions at once.
Autonomous trading creates a different transmission channel. Trading firms currently use AI agents mainly for lower-risk activities such as research, but systems responding to the same signals could intensify herd behaviour during periods of stress. The Bank of England is working with the BIS Innovation Hub and Deutsche Bundesbank on simulations designed to identify which features of agent design could drive common reactions. Possible controls include market-wide circuit breakers or kill switches if faulty models begin to destabilise trading.
Agent-led payments raise questions over consent, liability and fraud. As technology companies, merchants and payment providers automate transactions, regulators will need to determine how customers authorise repeated purchases, how disputed payments are resolved and who bears responsibility when an AI agent makes an erroneous or fraudulent transaction. Existing technology-neutral rules may require revision if human approval is no longer present at every stage.
The ECB forum showed that central banks now view AI through two financial channels: the productivity and efficiency gains it may create, and the systemic exposure that can build through debt, concentrated technology providers and automated decision-making. The Bank of England’s updated Financial Policy Committee assessment, due on July 7, will provide the next indication of how those risks may be reflected in supervision.
Finance teams adopting AI will need a clear record of model ownership, funding commitments, third-party dependencies and recovery arrangements. The strongest business case will be one that links efficiency gains to tested controls and credible continuity plans before autonomous systems become embedded in trading, payments or core financial operations.
More from Finance Monthly - Bain Capital Raises Kakaku Bid | Finance Monthly
