The AI advice gap: what happens when the machine says "yes"?

When seeking investment advice from a UK bank, individuals receive a lawful recommendation with the option to claim compensation through the Financial Services Compensation Scheme if negligence occurs. However, when asking AI chatbots like ChatGPT or Claude the same question, there is no such guarantee. Despite this, over a quarter of UK consumers now trust AI chatbots for financial guidance.

The Financial Conduct Authority (FCA) has highlighted this growing concern in its recent review, with other surveys reporting even higher usage rates. People are increasingly relying on AI for their financial decisions, yet the debate surrounding this trend has focused on accuracy rather than the more pressing issue: accountability.

Studies comparing chatbot advice to certified financial planners have shown that the AI consistently misses emotional and situational context, resulting in recommendations that are oversimplified and potentially misguided. In one case, ChatGPT incorrectly described Binance as FCA-registered when advising on cryptocurrency, despite Binance having been ordered to cease UK-regulated activity back in 2021.

While the AI's misleading outputs are concerning, the real problem arises when the incorrect advice has already been acted upon and gone wrong. In such cases, nobody takes responsibility, as AI platforms like ChatGPT and Claude are not subject to FCA regulation. Consequently, consumers using these chatbots for financial guidance have no access to the Financial Ombudsman Service or the Financial Services Compensation Scheme should things go wrong.

The regulatory gap exists because UK financial regulation is activity-based, not technology-based. A firm that provides personalized recommendations is engaged in regulated activity, regardless of whether those recommendations are provided by a person or an algorithm. AI chatbots occupy a new space, as they are not marketed as financial advisers and don't claim to be regulated.

This ambiguity, as addressed in the FCA's Mills Review, creates a significant gap between the actual financial impact and the protection provided by regulatory authorities. The review recommends that the FCA revise the scope for receiving financial recommendations developed using AI tools. In the meantime, fintechs can fill the void by separating information from personalization, ensuring their AI assistants clearly distinguish between general information and personalized recommendations.

Furthermore, fintechs should ensure their disclaimers carry weight, as merely adding "not financial advice" to recommendations won't suffice for regulatory compliance. While there is currently no law protecting ordinary people, fintechs should start building accountable versions of their businesses, incorporating human oversight and audit trails to ensure someone is ultimately responsible for the AI-generated advice.