In a speech on 19 October 2026, Jessica Rusu, the UK Financial Conduct Authority's chief data, information and intelligence officer, said firms' incident reporting should cover AI near-misses and that agentic systems need assurance that continues after deployment. It was a speech at Bloomberg's UK Policy Series, not a change to FCA rules.

What the FCA Said

From the published text, which the FCA notes is the drafted speech and may differ from the delivered version:

  • "Our rules already require firms to report major incidents, and that must extend to significant AI failures and near-misses."
  • "A system that is safe at deployment may behave differently at scale, under stress or as the models beneath it change. Agentic systems therefore require continuous, conditional assurance."
  • "The FCA will work with industry, fellow regulators and frontier labs to develop a trusted framework for agentic finance, including clear approaches to agent identity, delegated authority, accountability and execution."

Rusu also described the FCA's own use of agents. It is building AI tools for the oversight of 60,000 anti-money laundering firms coming into its remit in 2028, and a "market first responder agent" that will supervise about 9,000 small wholesale markets transaction reporting firms. She said the FCA launched its Handbook API in August, giving firms access to its machine-readable rules, and is working on machine-readable data collection APIs.

What It Means For Companies Running Agents

A company selling agents to UK financial firms can expect those firms to ask for two things the speech names: a record of near-misses as well as incidents, and evidence that checks are rerun when the underlying model changes. Neither can be produced after the fact for months that were not logged.

The four topics named for the planned framework, agent identity, delegated authority, accountability and execution, are the questions a firm will ask about any agent that acts for its customers.

What This Does Not Settle

  • The speech does not define a near-miss or a reporting threshold.
  • It gives no date or consultation for the agentic finance framework.
  • Any change to reporting rules would need a formal FCA process.

Related: The agent overreach containment playbook covers logging and containment, and the agent identity playbook covers identity and delegated authority. See also our notes on the ICO call for evidence on agentic AI.