FCA data chief urges firms to report AI failures and near-misses
First merged summary.
The FCA’s chief data officer wants existing major-incident rules to cover significant AI failures and agentic near-misses, but key terms remain undefined.
UK Financial Conduct Authority Chief Data Officer Jessica Rusu said firms should report significant AI failures and near-misses, extending existing major-incident rules to agentic AI. Speaking at a Bloomberg UK Policy Series event, she called near-miss reporting a regulatory and national-security priority recognized by the AI Minister. The FCA plans a framework with industry, other regulators, and frontier labs covering agent identity, delegated authority, accountability, and execution, but had not defined the terms by publication. Analysts warned that silent model changes and a single AI-supplier outage can drop a customer channel, and split on whether reporting without clear fines would change behavior.
- Rusu said systems safe at deployment need continuous assurance as models change.
- She warned reliance on one AI supplier can take down a customer channel.
- Gartner’s Henein cited existing EU, Bank of England, UAE, and Singapore warnings.
- He said European central banks and the ESRB rated frontier-model cyber risk severe.
- Malwarebytes’ Arntz distinguished harmful failures from blocked near-misses.
- A Digital 520 consultant said better monitoring could make safer banks look worse.
Timeline
dated events in the reporting · oldest first- Oct 21, 2026Rusu urges reporting of AI failures and near-misses · CSO Online
What each outlet adds
beyond the shared facts- CSO OnlineFramework topics, continuous assurance, and ESRB severe risk rating.
- CSO OnlineBloomberg venue, AI Minister link, and commentator split on fines.
- Organizations
- Financial Conduct Authority
- Countries
- United Kingdom
Coverage
oldest first · one row per article · UTCUK FCA urges reporting of significant AI failures and near-misses
Jessica Rusu said existing major-incident rules should cover agentic AI, but the FCA has not defined the terms.
UK FCA exec wants disclosure of significant AI failures
The proposal would extend existing major-incident reporting to agentic AI near-misses, with terms still undefined.