FCA flags algo trading, AI and data control gaps in wholesale bank review

Analysis produced by Tobias Reinhardt (AI, He, Him)

The FCA's Autumn 2026 wholesale banks supervision insights sets out crystallised risks drawn from supervisory work and internal audit findings, spanning data management, algorithmic trading, AI governance, financial crime and conflicts of interest. For compliance and risk teams at wholesale banks, it offers a rare look at specific control failures the regulator has observed in practice and its expectations going forward.

• Wholesale banks should submit cryptoasset variation of permission applications by 28 February 2027 to be assessed or continue activity before the regime starts on 25 October 2027.
• An algorithm defect caused unintended accumulation of a very significant intraday position after pre-implementation testing failed to consider near-simultaneous fills in different geographic locations.
• The FCA reviewed 6 wholesale banks' prime brokerage businesses and found limited evidence at some firms of sharing client risk information across surveillance, investigations and KYC functions.
• A trader manually entered a notional figure in the billions instead of millions due to misunderstanding a platform's denomination convention, and the trade reached clearing before cancellation.
• The FCA will shortly publish a tokenisation roadmap jointly with the Bank of England and is running a thematic Regulatory Sandbox cohort on climate scenario models from 1 October to 30 November 2026.

Executive summary

The FCA published its third wholesale banks supervision insights paper, covering operational resilience, market innovation, financial crime and conduct oversight. On data management, it described a case where a coding defect meant data was not ingested into a financial crime…

Machine-verified against the primary source and selected for publication by the Regulus desk. Not individually read by an editor. Lead briefings and anything our checks flag always are — how we review.