FCA publishes good and poor AI practice, asks AI providers for disclaimers

FCA chief data, information and intelligence officer Jessica Rusu said in a drafted speech at Bloomberg's UK Policy Series, dated Oct. 19, 2026, that the FCA is publishing good and poor AI practice drawn from its firm testing this year and is calling on AI providers to use clear, proportionate and consistent disclaimers on financial information and advice generated by their models; the FCA noted the drafted text may differ from the delivered version.1

Rusu said the practice examples are meant to give firms practical guidance and examples, and that the FCA wants the UK at the forefront of safe adoption of AI in financial services.1 On disclaimers, she said the FCA does not want to expand its remit but to make the existing boundary clearer, and that warnings may work better where AI-generated financial information is delivered.1

As good practice, Rusu cited one firm that built a live list of vulnerability signals its chatbot uses to flag consumers who may need extra support to human advisers; she also cited a firm with an independent ethics panel that reviews AI use cases before deployment.1

As poor practice, she listed AI tools built without a clearly defined need, vulnerable consumers excluded by default, overstated capabilities, automated customer journeys deployed without sufficient testing, and governance added after deployment.1

Rusu said firms need senior managers personally accountable for AI systems, and said firms should understand supplier concentrations, including reliance on a single model.1

She said only 40% of consumers correctly recognise that there is no recourse if something goes wrong after acting on advice from a general-purpose AI tool, and named ChatGPT, Gemini and Claude as examples of such tools; the speech text available does not say whether any provider has agreed to the request.1