UK FCA Warns of Financial AI Arms Race, Calls for Oversight of Large Models

General-purpose AI tools are increasingly being used to bridge Britain’s financial advice gap by helping consumers digest complex investment information at little or no cost. But services such as ChatGPT and Gemini are not authorised financial advisers, and their outputs may be inaccurate or outdated. Advice from these tools generally falls outside Financial Conduct Authority regulation, leaving users without the complaint and compensation protections attached to regulated advice.

Research released by the FCA on Aug. 27, 2026, covering 18- to 40-year-olds who own or are considering investments found that 44% mistakenly believed AI-generated financial information was regulated. Some 38% said it was acceptable to make an investment decision solely from AI output, while 32% wrongly expected recourse through the Financial Services Compensation Scheme or the Financial Ombudsman Service. Four in five less-experienced investors had used AI for investing help.

Britain is preparing to centralise anti-money-laundering and counter-terrorist-financing oversight of legal, accounting, and trust and company service providers at the Financial Conduct Authority. The work is now spread across 25 professional body supervisors overseen by the FCA’s OPBAS unit. The government’s 2025 decision is significant because it seeks more consistent enforcement across professional services and will sharply expand the FCA’s workload, making scalable data analysis and automated document review central to the new supervisory model.

The FCA said in its March 26, 2026 work programme that it expects to supervise about 60,000 additional firms, versus roughly 35,500 firms under its current broader remit. It plans to roll generative-AI document review into authorisations and supervision after successful testing. On July 15, Chief Executive Nikhil Rathi told lawmakers the watchdog was already testing agentic market surveillance to process billions of data rows each day, back-test records and seek corrections automatically, while retaining human accountability and auditable decision trails.

Artificial intelligence has been used in banking for years, but generative and agentic systems are shifting it from a support tool toward an operating layer for retail finance. The Financial Conduct Authority’s Mills Review, led by Executive Director Sheldon Mills, examines how that transition could reshape firms, consumer decisions and competition through 2030 and beyond. The stakes extend beyond innovation: greater autonomy can blur accountability, concentrate market power and magnify fraud, cyber and consumer-protection risks.

The FCA published the review on July 6, 2026, identifying four system shifts and issuing seven recommendations. It found the UK’s principles- and outcomes-based framework remains fit for purpose, while urging the regulator to secure the regulatory perimeter, expand its AI Lab and build agentic supervision. An April survey of 5,026 retail-finance consumers found 20% — about 11 million UK adults — would likely use AI able to act autonomously within preset goals, underscoring pressure to clarify oversight before adoption accelerates.