Britain’s Financial Conduct Authority faces calls to examine whether large language models such as ChatGPT, Claude and Gemini need regulatory oversight as their role in consumer financial decisions expands.
A review commissioned by the FCA found that more than one-quarter of UK consumers trust generative AI tools for financial advice.
Many users do not understand that safeguards for regulated financial services do not apply to AI platforms, according to Beinsure.
Sheldon Mills, the FCA’s executive director and author of the review, said chatbots risk crossing the boundary between generic financial guidance and regulated advice. Personalised recommendations or continuous adaptive prompts could resemble advice reserved for authorised firms.
Financial advice remains a regulated activity in the UK. AI systems should provide general information rather than individual recommendations, yet the distinction isn’t always clean when a model responds to detailed personal questions.
Mills urged the FCA to review the scale, nature and market effect of AI models outside its existing perimeter within three to six months. The regulator should consider how it might secure and adapt its rules as financial firms and consumers rely more heavily on these systems.
Jonathan Herbst, global head of financial services at Norton Rose Fulbright, said Mills had not called for an immediate enforcement push. He said policymakers instead face a larger question over whether current rules fit the ways firms now deliver financial services.
The FCA described itself as the first regulator worldwide to study AI’s impact across financial services. It does not need to adopt Mills’ recommendations.
The review also addresses concentration risk. A recent survey found 81% of financial firms worldwide use AI at some level, while 40% have moved into more advanced deployment.
Most firms still use AI in lower-risk back-office work. British financial companies increasingly use it in customer-facing activity, including complaints handling and investment guidance.
Mills warned that widespread adoption could leave firms dependent on a small group of model developers, cloud providers and technology suppliers. Shared systems could produce correlated behaviour, herd-like responses and common operational failures across the sector.
According to Beinsure, this issue moves beyond consumer protection. Financial institutions using similar models and infrastructure may expose the wider market to shared technology failures during periods of stress.
Sarah Breeden, deputy governor of the Bank of England, raised similar concerns last week. She said current regulatory frameworks did not anticipate autonomous AI agents, and firms cannot realistically rely on human approval for every action those systems take.
Breeden’s comments suggest British authorities are moving towards more specific AI rules for finance. The immediate issue centres on whether existing oversight can keep pace with systems that guide customers, automate decisions and depend on a limited group of providers.
