A column-secured or sensitive column indexed for AI retrieval
What it is
Sensitive column is indexed as AI knowledge.
Why it matters
A column you deliberately gated, with column security or by pinning it as sensitive in Pathix, is listed in a Dataverse knowledge configuration as indexed for retrieval, which puts it inside the set an agent or an app copilot is set up to answer questions from.
Pathix reports the configuration and nothing beyond it. The identity each consumer queries under is not read at this tier, so this never says a value was returned to anyone; it says the column was put in scope, which is the decision worth reviewing. High when an agent using the configuration authenticates nobody, Medium otherwise, including the app-copilot case where no agent exists. A table grounded with no per-column detail is never flagged, because the platform has not said which columns it exposes.
Find it yourself
Read dvtablesearch, one row per knowledge configuration, keeping active rows only, then its dvtablesearchentity children and the dvtablesearchattribute rows that list indexed columns. Cross that column list against column security in the table metadata and against your own sensitive list. A row carrying an app unique name and no bot is a model-driven app copilot. A table with no attribute rows is grounded at table level and cannot be checked per column.
How to fix it
Open the knowledge or generative-answers configuration and remove the column from the indexed set, or the table if none of its columns should be answerable; if it belongs there, confirm the audience: the agent's authentication and access settings, or the app's security roles.
No control mapping, deliberately
This is a security finding that carries no SOC 2, ISO 27001, NIST 800-53 or CMMC reference. That is a decision rather than an omission. Pathix maps a finding to a control only where the mapping is defensible to an assessor, and a stretched one would undermine every mapping that is real.
Pathix checks this across every environment you scan, along with 71 other conditions. Self-hosted in your own Azure, read-only, metadata-only.