The Guidelines scale with the FI: application is proportionate to size and risk profile. Except where the low-impact route in 2.3 is available, two things are expected. First, oversight: frameworks, policies and procedures governing AI use that include a clear way of recognising AI use cases, a sound assessment of how material each one is, and a proper inventory, as Sections 3 and 4 describe, with the breadth of those arrangements matched to the FI's activities, its AI use and its risk profile. Second, execution: the life cycle controls, capabilities and technology infrastructure of Sections 5 and 6, each applied where it is relevant to the use case, system or model concerned and implemented according to how material that use case is (controls on transparency, explainability and fairness matter most for credit, underwriting, advice and fund management, while data, safety and cyber controls always apply at a proportionate level).
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.