Britain’s Financial Conduct Authority (FCA) expects firms to report significant failures and near misses involving artificial intelligence, as they already must for other major incidents, a senior official said in a speech Monday. The remarks by Jessica Rusu, the FCA’s chief data, information and intelligence officer, set out the regulator’s position rather than a new rule, but they tell banks, insurers and investment firms operating in the U.K. to treat AI malfunctions as operational incidents with named executives answerable for them. The Register reported the speech Tuesday.
What Happened
Rusu spoke at Bloomberg’s UK Policy Series. In the text published by the FCA, which notes it is the speech as drafted and may differ from what was delivered, she said: “Our rules already require firms to report major incidents, and that must extend to significant AI failures and near-misses.” A near miss is a case where an AI system nearly caused harm but was caught in time.
She went further on agentic AI, meaning systems that can take actions on a user’s behalf rather than only answer questions, saying near-miss reporting for such systems “is becoming a national security priority,” a point she credited to the U.K.’s AI minister. She called for human oversight and for senior managers to be personally accountable for AI systems, and criticized firms that add governance only after technology is deployed, calling that “the wrong way around.”
The FCA plans to work with industry, other regulators and frontier AI labs on a “trusted framework for agentic finance,” including how AI agents are identified. Rusu said such systems need “continuous, conditional assurance,” because they may behave differently over time or at scale. The regulator is also working directly with banks on how frontier models are changing cyber defense, and she urged firms to make sure they hold proper cyber insurance.
She previewed a joint FCA and Bank of England survey due the following week, which she said shows that 86% of firms now use AI. She also said four in five less-experienced investors aged 18 to 40 have used AI to help with investing, and that only 40% of consumers correctly recognize they have no recourse if they act on general-purpose AI advice that goes wrong. The FCA is asking AI providers to put clear, consistent disclaimers on financial information their models generate.
Why It Matters
The speech did not announce a consultation, deadline, threshold or reporting format. Even so, the direction is clear for boards. Reporting near misses requires firms to detect them, which means logging, monitoring and escalation for AI agents before regulators ask for the data. Pairing that with personal accountability for senior managers puts AI risk on the same footing as other operational failures in regulated finance.
The emphasis on agents also matters. Rusu’s call for assurance that continues after deployment suggests a one-time model review at launch will not satisfy the regulator for systems that act on their own.
Context
The speech follows the government’s announcement last week that it will explore using the Cyber Security and Resilience Bill to require reporting of serious AI incidents and near misses, The Primary Wire reported, citing a government release. The government has yet to commit to adding AI-specific measures to the bill, The Register reported Tuesday. An official told members of Parliament last week that amendments remain under consideration, The Primary Wire reported, citing The Register.
What to Watch
- Publication of the joint FCA and Bank of England AI survey, which the speech said would come out the week after it was given, and the detail behind the 86% figure.
- Any FCA consultation or guidance that defines what counts as a significant AI failure or near miss, and how and when firms must report one.
- Progress on the agentic finance framework with industry and frontier labs.
- Whether the government adds AI incident reporting to the Cyber Security and Resilience Bill.
