Redact PII before sending prompts, code, or documents to ChatGPT, Claude, Gemini or any LLM

No. All detection and masking happens locally in your browser's memory. You can verify this yourself by opening your browser's Network tab or disconnecting from the internet — the tool keeps working.

ChatGPT, Claude, Gemini, Copilot, Grok, Llama, Perplexity, Mistral, and Hugging Face Chat. The browser extension adds in-page masking directly inside each of these.

The free tier covers core PII types (names, emails, phone numbers) up to 15,000 characters per scrub, with a 3-page PDF limit. Batch processing, OCR, and industry-specific detection profiles (legal, medical, finance, etc.) require PRO.

PrivacyScrubber is designed to support de-identification under HIPAA Safe Harbor, GDPR Article 25/32, and SOC 2 CC6-series controls by removing identifiers before data reaches a third-party AI tool. It is a technical control, not a substitute for a full compliance program — check with your compliance team on whether it fully satisfies your specific obligations.

Yes. Paste the AI's response back into the tool and click "Reveal" — it swaps the tokens back to your real data locally, with no server round-trip.

PrivacyScrubber is a 100% local, browser-based PII scrubber. Unlike cloud DLP solutions, it performs enterprise-grade PII and PHI scrubbing entirely in your browser's RAM, ensuring zero-trust data sanitization for texts and documents (PDFs, DOCX, CSV) without ever sending data to an external server.

The session map is stored exclusively in volatile browser RAM. The moment you close the tab, reload the page, or navigate away, the entire mapping — including all original PII values — is permanently destroyed.

Yes. PrivacyScrubber detects and scrubs PII from local documents (.txt, .docx, .pdf, .xlsx, .csv, and images). You can download the output either as a fully sanitized, anonymized document (retaining its original file extension and structure) or copy the processed text stream directly to your clipboard.

PrivacyScrubber generates an AES-GCM encrypted link locally. Your colleague can paste the AI's response and use your shared, encrypted key to safely 'reveal' the original data on their machine — no database or server involved.

PrivacyScrubber encloses both your editor input (#inputText) and the highlighted token output within isolated, closed Shadow DOM roots. This makes the raw, unredacted text completely inaccessible to any third-party scripts, browser extensions, or key-sniffing scripts running in your main document pool, creating a hardened frontend sandbox.

Yes. If your corporate credit card is declined due to international banking restrictions or if your procurement department requires manual invoicing, purchase orders (PO), or bank wire transfers (SWIFT/SEPA), our team can manually issue a formal invoice with tax reverse-charge support. Contact support@privacyscrubber.com to request an enterprise invoice.

To prevent PII or case documents from remaining in your device's memory after sanitization, the processing engine actively overwrites all temporary parsing buffers (used for DOCX, PDF, and spreadsheet extraction) at the byte level using Uint8Array.fill(0) in local memory immediately after use, ensuring zero sensitive residue remains in RAM before garbage collection.

No. Changing your detection profile (e.g., from General PII to Wealth Management) does not clear your active session or reset your local token maps. This ensures that any tokens generated using your previous profile remain fully decryptable when you paste AI responses back to 'Reveal' the original text. To completely wipe your session, rotate encryption keys, and clear all memory, click the 'Clear' button at any time.

PrivacyScrubber operates entirely in your browser's local RAM. Compliance audit receipts and team handoffs are exclusive features of the Teams Subscription. For audit compliance, the system builds an offline cryptographic PDF receipt documenting entity counts, timestamps, and active compliance profiles. For Teams Handoff, it encrypts the tab's session map in-memory using Argon2id and XChaCha20-Poly1305. Your colleague decrypts the token map on their machine using the shared key — no central database or server ever touches your data.

No. To maintain a strict zero-trust data sanitization architecture, PrivacyScrubber does not load third-party analytics trackers, cookies, or reviews integrations (such as Google Customer Reviews or similar widgets). Adding dynamic third-party scripts would require relaxing our Content Security Policy (CSP) and could allow external servers to inspect page states or log user environments. We display reviews statically or link to external platforms like G2 or the Chrome Web Store to ensure zero data leakage.

Yes. PrivacyScrubber offers an official local Model Context Protocol (MCP) server package (@privacyscrubber/mcp-server) that integrates directly with Cursor IDE, Windsurf, and Claude Desktop. It acts as a local prompt security gateway, scanning and masking PII and API keys locally in volatile RAM before they are transmitted to external LLM APIs.

In our Zero-Trust Data Sanitization (ZTDS) architecture, there are no server-side accounts or traditional 'Admin' vs 'User' roles. Team Administration is handled via client-side Blueprints. A team lead sets up the required compliance profiles, custom regex rules, and organization name locally, then locks the configuration and exports a cryptographic 'Team Blueprint' URL. When workers click this URL, their local PrivacyScrubber instance activates the TEAMS license, imports the mandated configuration, and disables their ability to edit or delete rules, ensuring organization-wide compliance securely without a backend database.

Custom Regex Rules and Token Labels serve distinct purposes in Zero-Trust AI sanitization. Custom Regex Rules are a pattern discovery engine used to identify new or proprietary identifiers (such as internal ticket numbers, project codes like PROJ-\d{4}, or custom account formats) that are not covered by default NER models. In contrast, Token Labels do not search for new text; instead, they rename the replacement placeholders for standard detected entities (for example, mapping NAME → PATIENT or ID → CASE_NUMBER) so that the LLM receives domain-accurate semantic context in its prompt without requiring custom regex code.

When prompts are sanitized using generic placeholders like [NAME_1] or [ID_1], LLMs may lose contextual role differentiation in multi-party prompts. Custom Token Labels allow PRO and TEAMS users to alias standard entity categories to domain-specific terminology (such as [PATIENT_1], [CLIENT_1], [PLAINTIFF_1], or [DEFENDANT_1]). This gives the LLM precise semantic clarity about each entity's functional role while preserving 100% anonymization and enabling seamless 1-click reverse-scrubbing ('Reveal') of the AI response.