NoeticGuard Shield

On-device PII and brand-term masking for AI chat sites — before prompts leave the browser.

NoeticGuard Shield masks sensitive text on your device, in Chrome, before it leaves AI chat websites. Local-first masking: the same @noeticguard/core engine used by our SDK runs inside the page. Your NoeticGuard workspace syncs policies, packs and quotas with a publishable key (ng_pub_…). Raw prompts are not uploaded to NoeticGuard for masking. What Shield does • Replaces emails, phone numbers, cards, national and tax IDs and other personal-data patterns with stable vault tokens such as [EMAIL_1] before you send. Works in English, Turkish or mixed chats. • Applies the add-ons your workspace has: Secrets Guard (API keys and tokens become [SECRET_n]), Brand Guard, Legal Privilege and COPPA / Student Data dictionaries, when purchased or granted. • Follows the policy toggles and pack opt-outs you set in the dashboard. • Media Guard (Startup and above) can block image and file uploads and image paste on supported chats. It blocks media. It does not read images or run OCR. • Sends usage counts to your dashboard so quota and audit views stay in sync. Telemetry carries counts and entity categories only. Never prompt text, cookies or secrets. • Optional in-product feedback lets your team report a missed value without pasting the raw data into a ticket. Supported sites Shield runs on major AI chat websites. The current list and setup notes are here: https://noeticguard.com/docs/browser-shield Limits • Detection is rule-based. It can miss values. Treat Shield as a technical control, not a guarantee. • It helps mask personal-data patterns that GDPR, KVKK and CCPA programs care about. Orientation only, not certification. • Private (Incognito) windows are skipped until you allow the extension there in Chrome. • Not a cloud DLP that rewrites prompts on our servers. Not a replacement for your LLM vendor. Not a desktop-app injector: managed desktop LLM egress is a separate IT path (see our docs). Getting started 1. Create a workspace at https://app.noeticguard.com. The Developer plan is free. 2. Create a publishable key (ng_pub_…) under API Keys. 3. Open the Shield popup, paste the key, and confirm "Connected". 4. Open a supported AI chat site and send a test value, such as a fake email or a made-up key. The value is marked with a shield icon, and the model works with a token such as [EMAIL_1] instead of the original. FAQ Q: Which sites are supported? A: See the current list and setup notes: https://noeticguard.com/docs/browser-shield Q: Do I need a paid plan? A: You can start on the free Developer plan. Some packs (Secrets, Brand, Media Guard, industry dictionaries) need a paid plan or an add-on. See Dashboard → Billing / Add-ons or https://noeticguard.com/#pricing Q: Will my team still understand the chat? A: Yes. Tokens are stable, and the original values are restored for you locally from the vault. The model receives tokens, not the original values. Q: Does Media Guard read images? A: No OCR. On Startup and above it can block uploads and image paste, so screenshots of secrets are less likely to leave the browser as files. Q: Does it work in Incognito? A: Only after you allow the extension in private windows in Chrome's extension settings. Q: What about desktop apps (not the browser)? A: Shield is the Chrome extension path. Managed desktop LLM egress is a separate IT track, not this listing. Q: Where can I learn what not to paste into AI tools? A: SecOps guides: https://noeticguard.com/guides Links Supported sites and setup: https://noeticguard.com/docs/browser-shield Guides: https://noeticguard.com/guides Privacy: https://noeticguard.com/privacy Security: https://noeticguard.com/security Support: support@noeticguard.com