Skip to content
Sanitized AI Help Center

Using the extension

Once installed and signed in, the extension works in the background. You only hear from it when a prompt needs your attention.

When a prompt is flagged

If a prompt contains something sensitive, the extension stops it before it’s sent and shows you what it found. The warning names the category it matched — an email address, an API key — so you know exactly what to fix.

You’ll see one of two warnings, depending on how your organization has set up that particular guardrail.

Security Check (a warning)

The usual case. You can:

  1. Sanitize Prompt — replaces the sensitive parts with safe placeholders. Review the result, then Insert Sanitized to swap it into your prompt.
  2. Edit Prompt — go back and reword it yourself.
  3. Send Anyway — send it unchanged.

Blocked by Policy

Some guardrails protect data your organization has decided must never reach an AI tool. For those, the banner reads Blocked by Policy and there is no Send Anyway button.

You can still:

  1. Sanitize Prompt — usually the fastest way through. Once the sensitive data is replaced, the prompt sends normally.
  2. Edit Prompt — remove or reword the flagged content yourself.

If you genuinely need to send it as-is, click Copy Details for Admin. That copies a short summary — which guardrail fired, what type of data, which tool, and the time — to paste into an email or chat message. It does not include your prompt text. Your admin can grant an exception if the use case warrants one.

Why this was flagged

Both banners explain why. Your admin may also have added a note of their own — for example, “Patient identifiers can never enter AI tools, per our privacy obligations.” If you’re unsure why something was flagged, that note is the fastest answer.

If you think the flag is wrong

Expand the detected items and click Report False Positive. You choose which items you disagree with and can add a comment. The flagged text and a little surrounding context go to your admins for review — your full prompt is not shared.

Icon states

  • Active (coloured): you’re signed in and protected.
  • Grey / inactive: you’re signed out, or your account isn’t in an active organization. Click the icon and sign in.

Supported tools

The extension protects prompts in web-based AI tools such as ChatGPT and Claude. Your admin can tell you which tools your organization has enabled.

Some organizations apply stricter rules on higher-risk tools, so the same prompt may warn on one tool and be blocked on another. That’s intentional.