UK regulator plans a framework for AI agents that act and pay on customers' behalf

The UK Financial Conduct Authority said on 19 October 2026 that it will develop a framework for agentic finance, meaning AI systems that act for customers, including managing payments and transacting with other agents. Jessica Rusu, the regulator's chief data, information and intelligence officer, set out the plan in a speech at Bloomberg's UK Policy Series. According to the speech, the framework will cover four areas: how agents are identified, what authority a customer has delegated to them, who is accountable when something goes wrong, and how transactions are executed. The FCA said it will build it together with firms, other regulators and frontier AI labs. Rusu also argued that existing duties to report major incidents should extend to significant AI failures and near misses, and that agentic systems need continuous assurance because a system that is safe at launch may behave differently at scale or under stress. On the same day the FCA published a good-practice paper on AI use in financial services. No consultation date or timetable for the agentic framework was given, and the speech did not set out payment-specific rules. The announcement matters to online merchants because delegated authority is the open question in agent-led checkout. When a shopper's AI assistant places an order, merchants, payment providers and card schemes still need to agree who authorised the payment and who carries the loss if it was wrong. Card networks have published their own agent frameworks in recent weeks. A UK regulatory view on identity and delegated authority would shape how authentication, liability and disputes are handled for agent-initiated payments to UK customers.