Why it matters
Public-safe knowledge is approved for customer-facing AI: no internal paths, draft language, employee data, or unreleased features. One unsafe chunk in a public index is an incident waiting for a screenshot.
How it works
Label collections public-safe after review. Block indexing of non-safe metadata. Run red-team queries attempting to elicit internal content.
Example
Nintendo public-safe set includes verified Help Center and Refund Policy mirrors. Internal escalation playbooks live in separate collections with hard deny on customer bot policy.
Common mistakes
- 1Public-safe labels applied at folder level without page review
- 2Safe collections referencing unsafe embeds
- 3No red-team retests after permission changes
Your AI does not need more access. It needs the right access.
Set policies that control what each AI application can read, cite, and answer from.
See AI access policies