Bank of England warns: autonomous AI agents can amplify market stress

Bank of England Deputy Governor Sarah Breardon stated that the use of autonomous AI agents "is able to amplify volatility in periods of stress" in financial markets, and suggested that stricter regulation may be needed. She made this statement at the annual symposium of the European Central Bank in Sintra, Portugal.

What exactly concerns the regulator

This is not about trading algorithms in general — they have existed in markets for decades — but specifically about autonomous AI agents: systems that independently make decisions about buying, selling, or reallocating assets with minimal human involvement. Breardon points to the risk that multiple such agents, trained on similar data and following similar logic, may act in sync during a moment of market stress — amplifying rather than smoothing price fluctuations.

The problem of "uniform reaction" is well known to financial regulators from the era of algorithmic trading: when dozens of market participants use similar models and react identically to the same news trigger, this can turn a localized selloff into a cascading crash. With AI agents, this risk is amplified by the fact that their decision-making logic is far less transparent than classical algorithmic trading rules, and their response speed is orders of magnitude faster than human traders.

"The use of autonomous AI agents can amplify volatility in periods of stress," said

Sarah Breardon.

Why regulators are talking about this now

Central banks and supervisory bodies have been monitoring the role of algorithmic trading in sharp crashes and volatility spikes — so-called "flash crashes" — for several years. The emergence of more autonomous AI agents capable of independently interpreting news and reassessing positions without trader involvement adds a new level of unpredictability: the speed at which such systems react to market events is far higher than human speed, and the decision-making logic is often opaque even to those who developed them.

• Statement made by Sarah Breardon, Deputy Governor of the Bank of England
• Platform — annual symposium of the European Central Bank in Sintra, Portugal
• The exact formulation of the risk — AI agents' ability to "amplify volatility in periods of stress"
• Breardon acknowledges the possible need for stricter regulation of autonomous AI agents in finance

How regulators may respond to this risk

Central banks have already accumulated experience in regulating algorithmic and high-frequency trading: automatic stop mechanisms, limits on order volumes, and requirements for algorithm testing before deployment are applied. It is reasonable to expect that similar tools — disclosure of decision-making logic, mandatory stress tests of models, and limits on the share of assets managed by autonomous agents — will become the starting point for discussion on regulating AI in finance. The challenge is that modern AI agents, unlike classical trading algorithms with clear rules, make decisions based on models whose logic is not always explainable even to developers, complicating traditional supervisory approaches.

The Sintra symposium is one of the key European platforms where central bankers annually exchange assessments of risks to the financial system, and the fact that the AI agent topic was raised there suggests that regulators view it not as hypothetical but as urgent.

What this means

The statement by one of the leaders of the Bank of England shows that the topic of autonomous AI agents has moved from technology news into the agenda of financial regulators: the next step may be specific requirements for banks and funds to control systems that make decisions without human involvement.