AI Red Team Scope

AI red teaming is only useful when its scope is explicit. This coverage matrix defines the surfaces that were actually exercised during testing so reviewers can distinguish real assurance from assumed coverage. It maps attack surfaces across prompt and instruction handling, retrieval and corpus behavior, agent orchestration, tools and MCP, identity and authority boundaries, multimodal input, and application workflow. It also distinguishes testing activities such as hypothesis formation, execution, reproduction, evidence collection, and retesting, so the report shows not just what was targeted but how thoroughly it was validated. Service coverage is marked by state, including fully included, partial, optional, and environment-dependent cases, with scope confirmation required where authority or deployment conditions change the result. The result is a compact, auditable view of what the red team actually tested, what remained partial or out of scope, and where findings should be interpreted with caution.