Consent and what data Shield Web collects

Shield Web is built consent-first: it collects nothing until you (or, in managed deployments, your organization) explicitly allow monitoring.

The consent prompt

On first run the popup shows a disclosure titled AI usage monitoring active with Allow monitoring and Not now buttons plus a link to the privacy policy. Until you choose Allow monitoring, no discovery or audit events are sent.

What is recorded — and what never is

With monitoring enabled, Shield records:

• which AI services you visit (the tool's name, e.g. ChatGPT),
• the type of action and its outcome (e.g. "a prompt was submitted", "blocked"),
• when on-device detection is enabled, category labels of sensitive data (e.g. "pii") and counts.

Events go only to your organization's own Shield Control workspace.

What is never collected: your prompts, pasted text, file contents, or the sensitive values themselves. Page content is read locally on your device for detection only — it is not transmitted.

Turning monitoring off

You can withdraw consent at any time: the AI usage monitoring toggle in the popup (and on the options page) turns all collection off. While it is off, the extension sends nothing.

Managed organizations

If your organization rolls Shield Web out via managed policy, it can enforce monitoring as the data controller. In that case no consent prompt appears, and the toggle shows as on and locked with the note that monitoring is enforced by your organization's policy. Questions about that decision go to your admin or works council — the technical guarantees above (no prompts, no content, no sensitive values) apply unchanged.