Autonomous technology could trade without human oversight, raising risk of market crash, UK officials warn
AI agents that can make decisions and carry out transactions independently could soon be operating in the financial system, the Bank of England has warned, adding that relying on human supervision may no longer be a reality.
Bank of England Deputy Governor Sarah Breeden issued the warning at the European Central Bank’s annual conference in Sintra on Tuesday, as major developers move from AI chatbots to agent systems.
Unlike AI productivity tools like ChatGPT, Gemini, and Claude, which generate text, images, and code in response to prompts, agent AI is designed to complete multi-step tasks with minimal human intervention.
Instead of just recommending investments or recommending purchases, an AI agent can trade, pay, book products and services without requiring human approval for each action.
Breeden said the financial system is possible “transform into one that operates more freely, at scale and at speed.”
As AI agents become more capable, “Relying on a human being in the loop for all agency actions is unlikely to be realistic,” He said, saying that the current control systems “they weren’t built to think” this situation.
Breeden warned that AI agents trained on the same data and responding to the same market signals could exhibit herd behavior, making similar trading decisions at the same time and amplifying market volatility during times of stress.
Regulators should investigate “if defenses are needed, similar to circuit breakers or kill switches that can block or stop trading in the entire market if bad AI models cause the market to crash,” Breeden said.
He also urged the central banks to prepare for more frequent “the wonder of technology,” saying recent advances in AI have already exceeded policymakers’ expectations.
His remarks come as governments treat AI as a matter of national security. Washington has recently intervened in the issuance of advanced AI models on cybersecurity issues, while AI developers continue to move forward with agent systems. Anthropic launched the Claude Sonnet 5 this week, its latest model designed to perform complex tasks on behalf of users.
The push toward AI for agents has also reached the Pentagon. Last week, it revealed AI agents designed to continue to analyze intelligence and provide targeting options to commanders, while insisting that humans will retain final decision-making authority.
You can share this story on social networks:
Follow RT on
