Create a policy

Open Policies and select Create policy. Enter a Policy name, then choose Data condition, AI destination, Endpoint group and Planned response. Save sample policy appears in demo mode; Save draft appears in the local setup workspace.

The data conditions are Personal data (PII), Confidential documents and Uninspected content. Destination can be All tools, ChatGPT, Claude or Kimi. The group selector uses the fictional endpoint groups in demo mode and groups from your local deployment plans in setup.

Understand sample matching

An enabled rule matches when its group, destination and data condition match a fictional event. Personal-data matching uses the sample's Personal data classification. Confidential-document matching uses Confidential or Highly confidential sensitivity labels. Uninspected-content matching uses the inspection state.

The card shows the number of sample matches and the rule revision. A label of Internal alone does not match the Confidential documents condition in this preview. Open an event's Evidence trail to see the enabled rules currently matching that event.

Edit, pause or remove a rule

Choose Edit to update the rule. Use the card's enable switch to pause or resume it. These edits increase the local revision; they do not send a policy to an agent. Remove policy opens a confirmation and removes the rule only from the current workspace.

If setup has no rules, Use starter policies adds three suggested drafts. The local workspace supports up to 100 policies. Demo enterprises and setup keep their rules separately.

Planned responses are not enforcement

Monitor, Review, Warn and Block describe planned behavior. In this release, all evaluation is local or fictional. Warn and Block do not interrupt an upload, notify an employee or change any endpoint. Treat enabled setup rules as drafts and use demo matches to review their scope before any future deployment.