Shadow AI Firewall 作者： Varveroglou Vasileios

Shadow AI Firewall helps reduce accidental exposure of personal and sensitive data when you use popular AI assistants (ChatGPT, Claude, Gemini, Copilot, Perplexity, and others matched by the add-on).

What it does
- Runs only on supported AI websites from the manifest.
- Intercepts outgoing requests so payloads are scrubbed before they are sent when the add-on detects patterns you enabled (for example emails, IBANs, payment-card-like numbers, Greek tax/social identifiers, and optional NLP-based masking for names/places).
- Keeps a local mapping so replies that reference placeholders can be shown with your original text in the page UI where the add-on applies unmasking.

What it is not
- Not a complete enterprise DLP suite; heuristics can miss items or flag false positives.
- Does not replace your organization’s security policy or legal review.

Privacy
- Default/personal use: settings are stored locally. Optional enterprise mode may send minimal metadata (e.g. scanner type and host class) only if you configure a key; see the privacy policy.

Support
Build and source submission notes for reviewers are in the repository README (Mozilla AMO section).