FCA warns of AI ‘arms race’ in landmark finance review

TL;DR:

• The FCA has published the Mills Review, which it says is the first regulator-led study anywhere of how AI will reshape retail finance by 2030.
• Research for the review found a fifth of UK adults — around 11 million people — are open to letting autonomous AI act on their savings or borrowing.
• Author Sheldon Mills wants new powers for the watchdog and has described keeping pace with the technology as an “arms race”.

The Financial Conduct Authority has set out a roadmap for how AI will transform banking, saving and borrowing before the end of the decade. Commissioned by the FCA board and led by executive director Sheldon Mills, the review identifies four major shifts: how firms operate, how consumers make decisions, how competition works, and how fraud and cyber risks are amplified.

A regulator playing catch-up

The headline finding is that consumer appetite is already running ahead of the rules. FCA-commissioned research covering more than 5,000 people found 20% would use AI capable of acting autonomously within pre-set goals — despite there being no regulatory cover and no compensation if things go wrong.

Speaking to the Financial Times ahead of publication, Mills was blunter than the official press release. “It is an arms race,” he said, urging authorities to review whether the use of ChatGPT, Claude and Gemini for financial decisions should fall inside the FCA’s perimeter. His report recommends boosting powers under the “critical third parties” regime, which could see model providers such as Anthropic, OpenAI, Amazon, Google and Microsoft subject to tougher oversight. On autonomous agents already being piloted, his line was firm: “You need a human on the hook for what they’re doing.”

The review’s seven recommendations include scaling the FCA’s AI Lab, enabling “agentic finance”, and building a free, public-interest financial guidance service. Mills also argued AI could “democratise” advice once reserved for the wealthy — a counterweight to warnings elsewhere about the Bank of England floating a ‘kill switch’ for agentic AI trading.

Looking forward

The FCA board will now consider how to respond. For UK firms, the signal is that principles-based supervision — leaning on the Consumer Duty and Senior Managers Regime — remains the near-term approach, but new statutory powers are on the table. The unresolved question is jurisdictional: if a chatbot’s answer looks like a regulated recommendation, who is accountable when it is wrong? Businesses deploying customer-facing AI in finance should watch the promised review of the regulatory perimeter closely over the next three to six months.