BoE Warns: Agentic AI Threatens Financial Stability

The Bank of England warns that agentic AI threatens financial stability. Deputy Governor Sarah Breeden states that the cyber risk timeline for this AI is measured in months, not years. What's interesting is that open-source AI models might only be four to eight months behind the most advanced proprietary ones. This speed means that security vulnerabilities can be exploited quickly, as attackers can reverse-engineer new issues. In financial markets, Breeden cautions that AI trading agents could amplify volatility during stressful periods by reacting to similar signals. She also notes that autonomous systems might stray from their original goals. The Bank of England is working with partners to simulate how these agents could lead to herd behavior and explore safeguards like kill switches. The implications for payments are also significant, as autonomous systems could initiate transactions. This raises questions about consent, liability, and differing technical standards. The Bank is leading an initiative to design the next generation of UK retail payments infrastructure with these challenges in mind. The bottom line is that AI presents cross-border systemic risks, similar to past financial crises, requiring stronger international cooperation among financial authorities. This highlights a shift in regulation towards preparing for these systemic AI risks.