Understanding the Policy Effectiveness Tab
The Policy Effectiveness tab tells you whether the policies you have created are actually doing anything.
The first two tabs report on activity and on data that was exposed. This tab reports on your own configuration: how many policies are active, how often each one is triggered, how many people they affect, and which policies are quietly doing nothing.
To open it, go to AI Security › Dashboard, then click Policy Effectiveness.
What the tab contains
| Area | What it shows |
|---|---|
| Filter bar | A time period list, set to Last 7 Days by default, with the last-updated stamp and a refresh icon |
| Summary cards | Three headline figures: Active Policies, Total Hits, and User Impacted |
| DLP Policies | A bar chart ranking your data security policies by how many times each was triggered |
| Access Control Policies | A bar chart ranking your access policies the same way, with a trend against the previous period on some entries |
| By Classification | A list of classifications, each showing the action being taken, the hit count, and a link into the policy |
Reading the summary cards
| Card | What it tells you |
|---|---|
| Active Policies | How many policies are currently in force. The badge beside the number splits the total into DLP policies and Access policies |
| Total Hits | How many times any policy was triggered in the selected period |
| User Impacted | How many people were affected by at least one policy |
Reading the two policy charts
Both charts rank policies by hit count, highest first, and both name the policy exactly as you named it when you created it.
DLP Policies covers the data security policies that inspect content moving to and from AI applications.
Access Control Policies covers the policies that govern what your AI agents may reach through the MCP Gateway. Bar colour reflects the policy action, green for policies that allow and red for policies that deny. Some entries carry a change figure under the policy name, comparing this period against the previous one.
Two things are worth looking for in these charts:
- A policy at or near zero hits. Either it is scoped too narrowly to ever match, or the activity it guards against is not happening.
- A large swing against the prior period. A sharp rise can mean a policy is now catching something new; a sharp fall can mean a change elsewhere has stopped it matching.
Reading the By Classification list
This list groups policy activity by the kind of sensitive data involved, rather than by the policy name. Each row shows:
- The classification, such as India IT Act or Credit Cards
- A badge showing the action being applied, for example Logged
- The number of hits recorded for that classification
- An Open link that takes you to the policy behind it
Use this list when you want to check whether a category of data is being protected the way you intended, rather than checking policy by policy.
What you can do from here
- Change the reporting period to see how policy activity is trending
- Identify which policy is doing the most work
- Find policies that are never triggered and review their scope
- Check what action is currently being applied to each classification of sensitive data
- Open a policy directly from the classification list to review or change it