Docs / Browser Shield

Browser Shield (Chrome extension)

For employees using ChatGPT, Claude, Gemini, and similar sites in Chrome. Shield runs the same core engine in the page before the browser POST leaves. Not a repo scanner — for that see CLI scan. Not for your product's end users — use the SDK.

On this page+

Setup

  1. Sign in at app.noeticguard.com
  2. Open API Keys → generate a publishable key (ng_pub_…). Copy it immediately — it is shown once.
  3. Install from the Chrome Web Store: Add to Chrome.
  4. Open the Shield popup, paste the publishable key, click Save.
  5. Wait for Connected. Without a key, masking stays off.
  6. Optional: if you use ChatGPT / Claude in a Chrome Incognito window, open chrome://extensions → NoeticGuard Shield → Details → enable Allow in Incognito. Chrome blocks extensions in private windows until you flip that toggle. Managed fleets should prefer disabling Incognito — see MDM · Incognito.
  7. Use ChatGPT / Claude as usual — PII is replaced with tokens like [EMAIL_1] before the request leaves the browser.

Policies & Media Guard

Detector toggles, Brand Guard dictionaries, and Shield enforcement live under Dashboard → Policies and sync via GET /v1/config. Media Guard is a Startup+ flag that blocks screenshot / file uploads to LLM sites in Shield — no OCR, no media bytes to NoeticGuard.

Related: MDM & Chrome Enterprise · Brand Guard · Why classic DLP fails on ChatGPT

Chat language

Shield masks regardless of the language you type in ChatGPT, Claude, Gemini, or other supported sites. Turkish, English, and mixed prompts all behave the same for core PII: emails, phone numbers, payment cards, and other policy-enabled detectors run on patterns in the text, not on the UI locale of the LLM.

When you enable TCKN under Policies → Identity Data, 11-digit Turkish national IDs are masked whether the surrounding prompt is Turkish or English. Brand Guard redacts terms you add to workspace dictionaries — use Turkish, English, or any mix your policy requires.