FCA — Mills Review: Long-term impact of AI on retail financial services

Tag: S-2026-01-27-fca-mills-review Type: report Author(s): Financial Conduct Authority (Review led by Sheldon Mills, Executive Director — Consumers and Competition) Date of source: 2026-01-27 (call for input launched). Sheldon Mills’ Supercharged Sandbox Showcase speech on the review: 28 Jan 2026. Call for input closed 24 Feb 2026. Date ingested: 2026-05-17 (consolidated from ~10 Open Brain restatements 18 Mar – 8 May 2026) Authority weight: high — UK conduct regulator’s own review. Raw file: Open Brain corpus snapshot at /_raw_sources/open-brain-2026-05-17-corpus.md. External URL: fca.org.uk (direct fetch was blocked by network allowlist across the capture period).

What it claims

The FCA launched a long-term review into AI and retail financial services, “designing for the unknown” (Mills’ framing). The review examines how AI — particularly agentic and autonomous AI systems — could reshape consumer interactions with financial services by 2030 and beyond. It explicitly tests whether existing regulatory levers — Consumer Duty, SM&CR, Operational Resilience, and the Critical Third Parties regime — are fit to manage AI-driven risks. The review centres on four firm-level evidence criteria: explainability, fairness, resilience, and accountability, which together are emerging as the FCA’s de-facto AI governance assurance lens. Recommendations are due to the FCA Board in summer 2026, with subsequent external publication. The FCA has consistently signalled it will not introduce AI-specific rules, instead relying on existing frameworks; future guidance on audit trails and explainability is expected by end of 2026.

Notable quotes

• “Designing for the unknown” — Sheldon Mills, Supercharged Sandbox Showcase, 28 Jan 2026 (title of speech; verbatim quote of the surrounding text was not captured).

What’s speculative vs. asserted

• Asserted: scope of the review, named tester frameworks, summer 2026 reporting timeline, FCA’s general “no separate rulebook” stance.
• Speculative: the four test criteria (explainability / fairness / resilience / accountability) as a de-facto assurance lens — this is Paul’s synthesis from multiple FCA artefacts rather than the FCA’s own published label. Treat as practitioner inference, not regulator language.

Topics this feeds

• FCA approach to AI
• Model Risk Management and Agentic AI — agentic AI is explicit scope.
• AI Governance Maturity Gap

Open questions raised

• Whether existing frameworks will be judged sufficient for agentic / multimodal AI in the final recommendations.
• What FCA AI Lab “good vs poor practice” examples (due late 2026) will look like in practice.
• Whether SM&CR will be recalibrated as a result (signalled separately at the “Getting firms fit to run” speech, 22 Apr 2026).

Ingestion note

Consolidates Open Brain thoughts dated 5/8, 5/6, 5/3, 4/30, 4/29, 4/26, 4/24, 4/23, 4/17, 4/9, 3/31, 3/30, 3/29, 3/24, 3/18. All ingestions used WebSearch fallback on fca.org.uk; direct fetch was blocked throughout the capture period.