Agentic AI accelerates cyber and market risks, Breeden warns
Bank of England Deputy Governor Sarah Breeden warned at the ECB Sintra Forum that agentic AI is rapidly expanding financial stability risks across cyber defense, trading, and retail payments. She called on central banks to modernize resilience frameworks and global cooperation.
From double speed to autonomous action
AI capabilities are accelerating beyond previous projections, with the double-time for completing complex software tasks shrinking from seven months in 2019 to four months in 2024.
Breeden highlighted that generative tools have evolved into autonomous agentic systems capable of chaining complex actions without human intervention.
This acceleration poses immediate risks in cybersecurity, where open-source models lag frontier capabilities by only four to eight months.
Disrupted institutions face heightened vulnerability if patches are not deployed rapidly across interconnected technology providers.
The Bank of England is collaborating with the FCA and HM Treasury to establish practical guidance for institutions navigating these frontier risks.
Simulating automated market herding
Beyond cybersecurity, autonomous agents threaten market stability and payment infrastructure.
In financial markets, identical AI prompts could trigger herd behavior and amplify stress volatility.
To address this, the Bank of England is experimenting with the BIS Innovation Hub and the Bundesbank on multi-agent market simulations.
In retail payments, central banks are designing next-generation systems to support autonomous machine transactions while establishing new regulatory frameworks for user consent.
Regulators are chasing a moving target
Breeden rightly notes that existing rules fail when direct human control vanishes.
However, central banks lag dangerously in building system-wide failover options for core infrastructure.
Without binding international standards, global financial stability will remain vulnerable to automated contagion.
Source: Sarah Breeden: Agents of change
IN: