Skip to content
Sanitized AI Help Center

Guarded sites

A guarded site is an AI tool that Sanitized AI protects, such as ChatGPT or Claude. The extension only checks prompts on guarded sites that are turned on. Admins manage them on the Guarded Sites page.

Read the page

The cards at the top summarize your coverage:

  • Guarded now: how many AI tools are protected, out of those available.
  • Average risk in use: the average risk score of the tools you turned on.
  • Not protected: tools that are available but turned off, and how many of them are critical risk.
  • Your own sites: tools added for your organization only.

Each AI tool has a card with its risk score, a status switch, and a Report card link. To find a tool, search by name or address, filter by source or status, or sort by risk. You can also switch between grid and table view.

Turn AI tools on or off

  • One tool: use the switch on its card.
  • Several tools: select them, and then click Enable all or Disable all.

The extension picks up changes within a few minutes.

Monitor a tool before you enforce

Each tool that’s turned on is either Enforcing or Monitoring:

  • Enforcing: your guardrails apply in full. People see warnings and blocks.
  • Monitoring: prompts to this tool are checked and recorded, but nobody sees a warning and nothing is blocked.

To switch, click Enforcing or Monitoring on the tool’s card. For several tools at once, select them and click Monitor only or Enforce.

To monitor every tool at once, including tools you turn on later, go to My Organization > Protection and turn on Monitor the whole organization. Turn it off to enforce again; each tool’s own setting then applies. If some tools or guardrails are still set to Monitor on their own pages, you’re asked whether to enforce them too.

Monitoring is a good way to learn how a newly approved tool is used before you start interrupting people on it. Monitored events appear on the Events page when you turn on Monitor only.

Risk report cards

Sanitized AI assesses popular AI tools and maintains a report card for each one. Click Report card on a tool’s card to open it. It shows:

  • A grade and a score out of 10. The higher the score, the riskier the tool: Low risk, Medium risk, High risk, or Critical risk.
  • Which edition was assessed, consumer or enterprise, and when.
  • Data handling: whether the tool trains on what you type, how long it keeps data, where it stores data, what controls it gives organizations, and which security certifications it has. Each factor is rated Favourable, Partial, or Adverse.
  • Legal exposure: the tool’s terms on ownership of outputs, indemnity, liability, confidentiality, and more.
  • Required mitigations: what to put in place if you allow the tool.

Report cards also drive risk-aware enforcement, which can turn warnings into blocks on riskier tools.

Ask for a new AI tool

If your team uses an AI tool that isn’t on the list, click Request Guarded Site. Enter the tool’s name and address, and optionally describe how your team uses it. Choose a Priority, and then click Send Request. The Sanitized AI team reviews each request.