Bank of England warns agentic AI threatens financial stability

Bank of England Deputy Governor Sarah Breeden has warned that rapidly advancing AI capabilities, particularly agentic AI systems capable of autonomously carrying out complex sequences of actions, pose growing risks to financial stability.

Breeden noted that open-source AI models may trail the most advanced proprietary models by only four to eight months. She warned that delays in applying security patches can allow attackers to reverse engineer newly disclosed vulnerabilities, echoing the Five Eyes cybersecurity agencies’ assessment that the relevant timeline for AI-enabled cyber threats is measured in months rather than years.

Turning to financial markets, Breeden warned that AI trading agents responding to similar prompts or market signals could reinforce one another during periods of stress, amplifying volatility. She also cautioned that autonomous systems could drift from their original objectives or from broader public policy goals.

She said the Bank of England is working with the Bank for International Settlements Innovation Hub and Germany’s Bundesbank to simulate how different agent designs could contribute to herd behaviour. The work also explores safeguards comparable to market circuit breakers or kill switches that could halt AI-driven trading if faulty models threatened financial stability.

Breeden also highlighted the implications of agentic AI for payments, where autonomous systems could increasingly initiate transactions on behalf of users. She said this raises questions about consent, authorisation, liability for erroneous payments and interoperability as different organisations develop competing technical standards. The Bank is leading a public-private initiative to design the next generation of UK retail payments infrastructure with these emerging use cases in mind.

Breeden concluded by calling for stronger international cooperation, arguing that AI presents cross-border systemic risks comparable to those exposed during the global financial crisis. She suggested that the shared technology dependencies underpinning advanced AI warrant closer international coordination among financial authorities.

Why does it matter?

The speech reflects a growing shift in financial regulation from focusing on AI adoption to preparing for systemic AI risks. By highlighting autonomous decision-making, cyber threats and market dynamics, the Bank of England is signalling that agentic AI presents challenges that extend beyond individual firms to the stability of the financial system as a whole.

It also illustrates how central banks are beginning to rethink financial infrastructure for an AI-enabled economy. Questions around autonomous payments, liability, market safeguards and international coordination suggest that existing regulatory frameworks may need to evolve as AI agents become more capable of acting independently across financial markets and payment systems.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!