Bank of England Warns Autonomous AI Trading Agents Could Crash Markets

Deputy Governor of the Bank of England Sarah Breeden warned that autonomous trading AI agents are capable of amplifying volatility in financial markets and triggering sharp declines if they start reacting simultaneously to the same signals in identical ways. She made the statement at the annual forum of the European Central Bank in Sintra, Portugal.

The danger of synchronized AI agents

According to Breeden, the main threat to markets is not a single crash, but a chain reaction. If many autonomous trading agents simultaneously sell or buy assets, responding to the same signal in the same way, a feedback loop emerges: price movement amplifies itself instead of dampening.

• Statement made by Sarah Breeden, Deputy Governor of the Bank of England
• Presentation took place at the annual forum of the European Central Bank in Sintra, Portugal
• The discussion concerns autonomous trading agents based on artificial intelligence
• Breeden suggests that regulators may need new rules for such systems

Do markets need new rules for AI agents?

Breeden describes a scenario familiar to regulators from the era of high-frequency trading: when many algorithms massively copy each other's behavior, the market loses stability. The difference is that modern AI agents act far more autonomously than previous trading algorithms — they are able to independently interpret news and data, rather than simply execute pre-written rules. This, according to Breeden, may require regulators to develop new approaches to overseeing financial markets.

The platform on which the warning was made is also indicative. The annual forum of the European Central Bank in Sintra is one of the main venues where central bank governors align positions on risks to financial stability, and the appearance of the AI agents topic on the agenda of such a forum is itself a signal: central banks have stopped viewing autonomous trading systems as a narrow technical issue and are beginning to discuss it at the level of macroprudential policy.

How AI agents differ from previous trading algorithms

Algorithmic trading has existed for more than a decade, and regulators have already encountered episodes where synchronized behavior of automated systems sharply crashed quotations for several minutes. But previous algorithms typically executed rigidly written rules. AI agents, which Breeden speaks of, are capable of independently making decisions based on analysis of news, reporting, and market data — which means their mass synchronized reaction to the same informational occasion can prove less predictable and sharper than in the case of classical algorithms.

This key difference explains why Breeden speaks precisely about new rules, rather than the application of existing oversight of algorithmic trading. Regulators can track the speed and volume of transactions, but it is more difficult for them to predict in advance exactly how many independent AI agents will interpret the same event — and therefore to anticipate the moment when their individually rational decisions accumulate into collective panic.

What this means

The warning came from one of Britain's key regulators and points to growing concern among central banks about the systemic risk of autonomous AI systems in finance. For now, it is a public warning rather than concrete measures, but the very fact of such a statement at the ECB forum suggests that the topic of autonomous trading agents is seriously entering the agenda of financial regulators — and in the long term could lead to separate oversight rules specifically for AI agents, not just algorithmic trading as a whole.