FCA warned against ‘cognitive surrender’ as it becomes ‘agentic regulator’

On 19 October 2026, the FCA revealed plans to become an “agentic regulator” by using artificial intelligence (AI) for multiple purposes, such as intervening more quickly when problems are identified and analysing activities of regulated companies.

These plans accompanied new AI guidance, entitled AI in financial services: good practice and areas for improvement, which sets out examples and guidance for using AI across consumer outcomes, governance, accountability and operational resilience.

Speaking at the Bloomberg UK Policy Series, Jessica Rusu, FCA chief data, information and intelligence officer, said: “The FCA itself is changing.

”Part of our work to become a smarter regulator means developing our future as an agentic regulator, using the same technologies reshaping our markets to make ourselves sharper, faster and more effective.”

However, the shift from the FCA raises questions about what place agentic AI should have in regulation.

As the financial services sector adopts and advances its use of AI, Jonathan Fong, head of AI and digital at the ABI, believes that “it makes sense that the regulator adapts to this as well”.

Fong told Insurance Times that if the FCA follows a “similar approach” to its guide on good practice and areas for improvement in its own AI adoption, “then becoming an agentic regulator is a rational step to keep pace with change”.

He continued: “However, just like in other businesses, there are risks with AI being used in regulation.

“As agentic AI becomes more powerful, autonomous decisions may happen faster than the existing monitoring and governance measures were designed to manage. That’s why any organisation considering AI should take a proportionate approach to its use and include human oversight and continuous monitoring, with clear accountability.”

Regulatory risk

Meanwhile, Matt Twist, vice president of sales for Europe, the Middle East and Africa at Earnix, said the FCA’s ambition to become an agentic regulator highlights a larger shift for the insurance industry.

He explained that AI is moving “beyond generating insights into influencing and executing decisions of financial and reputational importance”.

An example of this is the FCA deploying automation and building AI agentic tools which will supervise around 9,000 small wholesale markets transaction reporting firms, so supervisors can prioritise more complex supervisory activities.

Nathan Wilmott, partner at Ashurst Perkins Coie, warned that the “FCA cannot allow AI to become in practice the decision-maker at any stage of the regulatory process.”

Wilmott continued: “The relationship of trust between regulator and the regulated community is equally important and this is something that cannot be delegated to machines.

“Similarly, the regulator’s statutory supervisory and enforcement powers must be exercised on the basis of human judgement based on all of the relevant circumstances. When increasing the FCA’s reliance on AI outputs by its policymakers, supervisors and enforcement staff, the regulator must put in place measures to guard against the very real risk of cognitive surrender to AI.”

Keeping pace

In response to the regulator’s increasing use of agentic AI, insurers will need to ensure their own governance processes can keep pace.

Kevin Gaut, chief technology officer at Instanda, explained that the regulator will be able to answer “better questions, faster” and “insurers will need to be ready to answer them”.

Gaut said that if an AI system has influenced an underwriting pricing or claims decision, “it will not be enough to point to the final outcome”.

This will be key to avoid the “uncomfortable scenario” where the regulator spots a pattern in customer outcomes before the insurer sees it in their own data, he said.

He continued: “Insurance already knows how to govern delegated decision-making.

“A cover holder is given defined authority, anything outside that authority is referred back, its activity is recorded and the insurer retains oversight. AI agents need much the same structure.

“The difference is scale. An AI agent could make or influence thousands of decisions very quickly and an AI-enabled regulator could analyse those decisions just as quickly. The audit trail therefore has to be built into the process. You cannot rely on reconstructing it afterwards.”

Twist added: “Ultimately, the measure of agentic AI should not be how much it can automate, but whether it helps insurers make better, more consistent and more timely decisions while maintaining appropriate control.

“Decision quality, business performance and customer outcomes must remain the measures of success.”