Data used anywhere in the AI life cycle must be suited to its purpose, representative, of high quality and properly governed. The FI's general data standards on ownership, access and intellectual property extend to AI data and are strengthened where AI needs more, and the FI checks that its existing data governance can cope with the volume and variety AI consumes. Seven areas matter: suitability of the data for the stated objective and context, judged alongside considerations such as fairness; whether training and test data represent the real conditions, stressed ones included, in which the AI will run; data quality in terms of relevance, accuracy, completeness and recency, with ongoing monitoring and checks for anomalies, drift and latent bias; classification that steers how data of a given criticality and sensitivity may be used; security across the data life cycle, including prompt destruction or sanitisation of training data, model artefacts and outputs once they are decommissioned or no longer needed, encryption in transit and at rest, and secure handling of the AI's inputs and outputs; privacy measures that follow the applicable rules and guidance, such as obtaining consent before sensitive personal data of customers or employees is used in training or processed by AI; and auditability and lineage, with records of where data came from, how it was selected and processed, how its suitability and representativeness were judged, and the approvals and remedial actions taken.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.