Pillar 1 of 9 · Data

How well do you control the proprietary data that reaches AI models?

Staff send whatever they like to public models. No classification, no gateway, no no-train clauses.

Policy exists on paper. No technical enforcement.

Central AI gateway with logging; no-train clauses in the main contracts.

Gateway plus automated redaction/DLP; documented residency; audit rights.

Above, plus the most sensitive classes never leave — in-VPC, on-prem or confidential compute.