Effective practices: Cyber risk and resilience management relating to artificial intelligence (AI)
RegCanary analysis of a PRA publication
Why it matters
RegCanary's analysis of what this means for regulated firms.
It does not introduce a new rulebook, but it highlights effective practices that supervisors are likely to test through existing cyber, operational resilience, outsourcing and governance expectations. For compliance teams, the central message is to treat AI as a critical capability rather than an isolated innovation project. Firms should be able to evidence board-level accountability, a complete AI use-case inventory, risk assessments covering model, data and third-party dependencies, and controls for model change, monitoring and incident response.
AI supply chains and cloud concentration need particular attention, along with data leakage, prompt manipulation, model opacity and the ability to contain AI-related incidents without disrupting critical services. Action is needed now: map AI use to critical operations, integrate AI scenarios into threat-led testing and business continuity, strengthen third-party assurance and contractual controls, and update policies, training and management information. Smaller firms can apply proportionality but should still document how they meet the same outcomes.
RegCanary expects supervisory dialogue to focus on evidence, not intent. Firms that move early can reduce disruption risk, improve regulatory confidence and use AI more safely in detection and resilience tools.
Who this affects
Sectors and firm types identified in the analysis.
- General Regulation
Source
PRA
Original publication, 19 November 2026
Read the original publication bankofengland.co.ukThis is RegCanary's analysis, not the regulator's text. It summarises and interprets the publication above; it is not the PRA's original wording, and it is not legal advice. Always rely on the original source.