“What these two examples – agentic commerce and agentic trading – both highlight is that, as AI capabilities increase, we must keep asking whether existing, technology-agnostic regulatory frameworks remain sufficient,” she said at a European Central Bank conference.

As companies increasingly adopt AI models, the use of agents is seen as one of the key ways the technology can help boost productivity by doing tasks on their own but guided by humans.

For consumers, this could mean AI booking holidays or refilling their fridges using the agents, according to Breeden. The finance sector could use agents to execute trading strategies.

“If AI agents respond similarly to the same prompts or triggers, they could amplify volatility in stress – especially if their objectives drift from original goals or public policy objectives, in a manifestation of the misalignment problem that can arise with some AI models,” Breeden said.

The UK central bank is working with the Bank for International Settlements and Bundesbank on understanding whether agents can drive “herding behavior” and how officials can tackle the problem. This includes whether to implement guardrails such as “circuit breakers or kill switches that would limit or stop trading market-wide if faulty AI models cause market meltdown.”

In a question-and-answer session, Breeden likened AI models to teenagers in that “they lie, they tell you they’ve not done things when they have and they behave differently when you’re watching them.” She added that it will be key for regulators to have a human who is “accountable for that model.”

BOE Governor Andrew Bailey has said the technology could help revive Britain’s tepid economic growth rates. However, UK officials are concerned about its effect on the banking sector, particularly if new models expose cyber vulnerabilities across the economy.

Breeden said international cooperation will be crucial, warning that new AI capabilities can spread across borders through common technology dependencies, globally systemic financial institutions and market infrastructure.

“What if the next surprise puts the latest capabilities in bad actors’ hands, or models develop in ways that are harder to evaluate and control, given evidence that some behave differently in testing compared to real-life scenarios?” she said. “We shouldn’t wait for a crisis to build the cooperation we need.”

— By Tom Rees (Bloomberg News)