The Financial Conduct Authority has told wholesale banks to strengthen controls around financial-crime data, artificial intelligence and electronic trading after its supervisory work exposed weaknesses capable of sitting unnoticed inside apparently mature systems.
One review found that a defect in data ingestion created a gap in financial-crime controls that remained undetected. The FCA’s point is structural: a control can appear to be operating while incomplete source data prevents it from testing the population it is meant to protect. Firms are expected to verify data lineage, completeness and reconciliation rather than rely only on the output of downstream monitoring.
The regulator also identified uneven AI governance. It pointed to incomplete inventories, conditional approvals that were not always followed through and uncertainty over whether firms could stop or withdraw a model quickly. Human accountability remains central even when a system is developed or supplied elsewhere.
Electronic-trading controls presented a third concern. The FCA said some safeguards against erroneous orders were insufficiently calibrated, leaving firms exposed to so-called fat-finger events and related market disruption. It expects limits and kill mechanisms to work in practice, not merely exist in policy documents.
These are supervisory findings, not a final enforcement decision against a named bank. The FCA’s publication also addresses the timetable for cryptoasset permission applications, giving firms a clearer route into the new regime. The official supervisory review was published on 2 November 2026.
