Frequently asked questions

It scans them — locally, inside your browser tab — but it never uploads them. The only thing that leaves your device is content-free audit metadata: detection type, action, domain, and timestamp. There is no code path that sends your text to our servers. See Privacy architecture.

ChatGPT, Claude, Gemini, Microsoft Copilot, and Perplexity have tuned profiles. Other AI-tool pages get a generic fallback that watches standard text inputs and send buttons. The full domain list is in the AI Guard guide.

No. AI Guard doesn’t scan on every keystroke — it scans at the moments that matter: when you paste, press Enter, or click Send. The scan runs on-device in milliseconds.

Warn pauses the send and shows you what was found — you decide whether to send anyway. Redact (Pro) replaces the sensitive values with placeholders and sends the cleaned text. Block stops the send until you remove the sensitive content.

Full on-device detection for PII, Secrets, and Financial data with the Warn and Block actions, plus the dashboard and content-free audit log. Pro adds the Redact action and the PHI and Source code categories. Compare plans in Billing.

Because there’s essentially no legitimate reason for a real card number or a live API key to reach an AI chat. Defaults are Warn for PII and PHI — where context matters — and Block for Financial and Secrets. You can change any of them.

Yes. Add custom keywords (project codenames, client names) and regex patterns (employee IDs, ticket formats) on the dashboard’s AI Guard page. They’re matched on-device like the built-in categories.

Yes — basic detection keeps running with your local settings. Signing in adds settings sync across devices, the audit log, and your plan tier. Core protection is never held hostage behind a login.

Yes. Detection is packaged inside the extension and the extension keeps a local copy of your configuration, so protection doesn’t depend on a network call. Audit events and settings sync catch up when you’re back online.

Not yet — ShieldWays is built for Chrome (Manifest V3). Other Chromium-based browsers may load the extension, but Chrome is the supported target.

Only partially, and we’re explicit about it: it redacts sensitive text visible in your meeting tab and masks sensitive input fields, but a browser extension cannot alter the outgoing camera or screen stream other participants receive. That requires a desktop app with a virtual camera, which is on our roadmap. Details in Meeting Guard.

Go to chrome://extensions, find ShieldWays, and click Remove. Your account and dashboard data remain; to delete those too, contact us and we’ll remove your account data.