In a speech, Jessica Rusu, chief data, information and intelligence officer at the UK Financial Conduct Authority, called on AI model providers to add clear, proportionate and consistent disclaimers when their tools give financial information or advice. The FCA published examples of good and poor practice from testing at firms. It said it will work with industry on a framework covering AI agent identity, delegated authority and accountability. It also stressed that major incident reporting must extend to significant AI failures and near-misses.

What it means: The speech shows how a major regulator is trying to get ahead of the shift from chatbots to agents acting on customers' behalf. The first message concerns the general-purpose AI models consumers use for investment questions. The FCA is not extending its remit to them, but it wants it to be clear that an AI answer does not carry the protection of regulated advice. The second message is for firms: they need accountable managers for each AI system, human oversight, stress testing and fallback plans if a model fails or changes. The intention to report near-misses as well is significant, because it means firms will have to monitor their agents' behaviour systematically. Finally, the FCA will itself use AI to supervise tens of thousands of firms. In the EU, similar issues are covered by the AI Act, DORA for ICT incidents and MiFID rules on investment advice, but rules for agents have not yet taken shape. For Greek fintechs designing autonomous services, the FCA's principles are a useful guide.

Read the original on FCA

Related service: myDATA integrations

Related stories