Every day, staff paste board minutes, patient records, source code, and M&A financials into ChatGPT, Claude, Gemini, and IDE coding agents. Blocking AI drives 60%+ underground Shadow AI; legacy per-seat DLP costs $30–$50/user and takes months to deploy. Built on the open ZTDS™ RFC v1.0 Standard, PrivacyScrubber sanitizes sensitive data 100% inside local endpoint RAM across all 5 corporate environments — giving CISOs un-bypassable Auto-Scrub governance, employees 1-click Encrypted Session Handoff, and engineering teams <0.4ms SDK pipelines for a single flat rate ($99/mo Unlimited Seats & Unlimited Volume).
Over 73% of enterprise employees paste client contracts, patient records, financial tables, and production source code into external AI tools (ChatGPT, Claude, Gemini, Copilot, DeepSeek) and IDE coding agents (Cursor, Claude Code, Cline, Roo Code). Rather than asking you to trust a proprietary cloud proxy with your secrets, PrivacyScrubber is engineered as the pioneer certified implementation of the open, vendor-neutral Zero-Trust Data Sanitization (ZTDS™ RFC v1.0) standard (IETF Internet-Draft draft-sibiryakov-ztds-protocol).
Every prompt, file, or database query is scanned and sanitized directly inside volatile endpoint RAM before a single network packet is created. Zero raw PII ever leaves the user's device.
Instead of static [REDACTED] bars that break AI reasoning, the standard substitutes numbered semantic tokens [NAME_1], [ORG_1], and [IBAN_1] so the LLM keeps 100% analytical precision.
When the AI replies using tokens, 1 click on Reveal restores real names and numbers locally on the employee's screen. Closing the browser tab immediately wipes the token map from RAM.
Because zero customer data transits a vendor server, no GDPR Art. 28 subprocessor DPA is needed, and every scrub generates a locally signed SHA-256 Compliance Receipt.
Establishes Statutory Risk Minimization & Technical Safe Harbor under GDPR Art. 25 & 32, HIPAA §164.514(b), SOC 2 CC9.1 & CCPA/CPRA.
Employees and developers use ChatGPT, Claude, Gemini, Copilot & Cursor at full speed — zero false-positive blocks and zero IT exception tickets.
Full protocol specs, academic benchmarks, and live Ed25519 certificates are publicly verifiable any time on ZTDS.ai.
When securing generative AI across the enterprise, CISOs, DPOs, and Legal Counsel face four immediate operational bottlenecks — each directly resolved by local ZTDS™ enforcement:
Firewall bans push staff onto personal phones and unmanaged accounts where security has zero visibility.
Raw PII/PHI entering external AI clouds exposes the enterprise and CISO/DPO to strict regulatory enforcement.
Rigid legacy DLP rules block legitimate prompts and flood IT security teams with daily exception requests.
Legacy DLP vendors charge per seat plus 6-month proxy rollouts, punishing company-wide adoption.
| Legacy Approach | Why It Fails the CISO in Practice | How PrivacyScrubber (ZTDS™ Standard) Solves It |
|---|---|---|
| Network / CASB DLP | Encrypted Traffic Blindness: Cannot inspect real-time browser AI chats or IDE streams without breaking TLS certificates and halting employee work. | Pre-Network Browser & IDE Shield: Scrubs sensitive data inside the endpoint before TLS encryption even happens. Zero proxy latency. |
| Cloud Scrubbing APIs | New Third-Party Breach Target: Sending raw prompts to an external DLP cloud requires a new GDPR Art. 28 DPA, vendor security audit, and adds 400–1200ms delay. | Zero Subprocessors (0-DPA): 100% of processing stays in local device RAM (<0.4ms). PrivacyScrubber never sees or touches your data. |
| AI Vendor "Opt-Out" | Zero Audit Proof: "Do not train" checkboxes do not stop subpoena exposure, vendor retention caches, or insider leaks, and give auditors zero proof. | Cryptographic Compliance Receipts: Generates a signed SHA-256 audit receipt for every session proving zero raw PII left the device. |
Whether an employee is chatting in a browser, scrubbing a 50-page PDF contract, sharing AI results with a manager, coding in Cursor, or running a backend RAG pipeline, the ZTDS™ Local Engine follows one simple loop in local RAM:
Employees never guess what lies behind a token. Hovering [ORG_1] opens the Spotlight Lens to preview its real value "Apex Global Holdings" in RAM without exposing other tokens.
If the engine masks a non-sensitive phrase like Project Horizon above, the user clicks × to restore it in 0.1s so the LLM keeps full context — zero IT tickets required.
Once sent, the AI reasons on [NAME_1] and [ORG_1]. When the AI replies, 1 click on Reveal swaps every token back to real names locally on the user's screen.
| Environment (Plane) | CISO Pain Solved & Simple How-To | Top Real-World Use Cases |
|---|---|---|
|
Plane 01 • Browser Shield
Chrome Extension & Auto-Scrub Shield ChatGPT, Claude, Gemini, Copilot, DeepSeek, Grok, Perplexity, Zendesk |
Pain: Staff paste client data into web AI chats. How-To: Shield sits inside the AI chat box. When CISO enables Auto-Scrub, every prompt is scrubbed automatically on send; 1 click on Reveal restores real names in the AI reply. |
• Support & Sales: Answering Zendesk/CRM tickets with AI without leaking customer PII. • Legal & M&A: Summarizing NDAs and term sheets in Claude. • Executives: Drafting board memos safely. |
|
Plane 02 • Offline Studio
Web Workspace & Batch Document Studio privacyscrubber.com (Included in TEAMS) |
Pain: Uploading full PDFs, Excel sheets, and scans to cloud AI. How-To: Drag-and-drop .pdf, .docx, .xlsx, .csv, or scanned images → click Scrub → download clean files + signed PDF Audit Receipt. |
• Healthcare: De-identifying patient EHR PDFs for clinical AI (HIPAA Safe Harbor). • Finance: Scrubbing multi-sheet Excel rent rolls, ledgers & W-2/1040 tax forms. • HR: Blind-screening candidate CVs. |
|
Plane 03 • Fleet Cockpit
Admin & Teams Governance Cockpit Teams Hub ($99/mo Flat Unlimited Seats) |
Pain: Shadow AI anarchy & slow DLP ticket queues. How-To: Admin locks mandatory industry profiles (Hard Lock) and shares a 1-click policy blueprint; workers add local deal codenames in 2 clicks + share sessions via Session Handoff. |
• 60-Second Fleet Rollout: Enforcing baseline rules across 50 to 5,000 employees via MDM or 1-click link. • Cross-Team Review: Analyst-to-Partner encrypted AI session handoff. |
|
Plane 04 • AI IDE Guard
AI IDE & Agent Firewall (MCP) @privacyscrubber/mcp-server |
Pain: Cursor, Claude Code & Cline reading .env keys and SQL dumps.How-To: Add 1 JSON block to IDE settings → agent file reads and DB queries are automatically sanitized locally before reaching the LLM. |
• Safe Vibecoding: Refactoring production repos without leaking AWS/Stripe keys. • Incident Response: Debugging live stack traces and SQL logs with AI agents. |
|
Plane 05 • Backend Engine
Developer Board & Backend SDK @privacyscrubber/sdk ($299/mo Flat Unlimited Nodes) |
Pain: Raw PII permanently polluting RAG vector DBs & microservices. How-To: npm install @privacyscrubber/sdk → call scrubText() before vector embedding or LLM API calls (<0.4ms in RAM, 0 network hops). |
• RAG Vector Pipelines: Scrubbing data before Pinecone / Weaviate / pgvector ingestion. • CI/CD Gates: Blocking secret/PII leaks in GitHub Actions (SARIF reports). |
Generic DLP tools flood teams with false positives because they do not understand industry context. PrivacyScrubber ships with 30 specialized industry profiles ready to activate in 1 click (all included in TEAMS $99/mo with unlimited characters):
| Sector Profile | What It Automatically Protects | Regulatory Safe Harbor |
|---|---|---|
| Healthcare, Clinical & Pharma | All 18 HIPAA identifiers, MRNs, NPI numbers, ICD-10 codes, clinical trial cohorts, and patient intake notes. | HIPAA Safe Harbor (45 CFR § 164.514), FDA 21 CFR Part 11. |
| Legal, M&A & Litigation | Case dockets, matter IDs, attorney-client privilege markers, M&A counterparty names, and settlement sums. | ABA Model Rules 1.6, FRE 502 Attorney-Client Privilege Protection. |
| Banking, Audit & Wealth Mgmt | IBAN, SWIFT/BIC, ABA routing, payment card numbers, portfolio AUM figures, and tax forms (W-2 / 1040). | GLBA, SOX Section 404, PCI-DSS v4.0, FINRA 4511. |
| DevSecOps & Cloud Engineering | AWS/GCP/Azure keys, JWTs, SSH private keys, Kubeconfigs, connection strings, and internal IPv4/IPv6 maps. | SOC 2 CC6.1, ISO 27001 A.8.11 & A.8.20. |
| HR, PeopleOps & Recruitment | Candidate CVs, compensation bands, performance reviews, National IDs / SSNs, and employee home addresses. | EEOC Blind Screening, GDPR Article 9. |
| Insurance, Real Estate & Gov | Policy & claim IDs, VINs, loss reserves, commercial rent rolls, defense CMMC / ITAR markings. | NAIC Model Law #668, NIST 800-171 / CMMC 2.0. |
Employees don't just paste text — they upload files. Here is how the built-in Offline Document Studio protects multi-format files inside local RAM:
Drop any contract, EHR chart, or legal brief PDF → sensitive text and hidden metadata layers are stripped locally, outputting a clean, shareable PDF.
Drop multi-sheet financial workbooks .xlsx or CRM exports .csv → every sheet and cell is tokenized in place while keeping formulas intact.
Drop photographed IDs, W-2s, invoices, or scanned lab reports PNG/JPG/TIFF → text is extracted and sanitized 100% on the device with zero cloud vision calls.
Engineering teams are the highest-velocity AI users — and the #1 source of accidental credential and production database leaks. PrivacyScrubber gives your DevBoard / Engineering Lead two drop-in packages that deploy in under 2 minutes with zero infrastructure changes:
Protects Cursor, Claude Desktop, Windsurf, Cline & Roo Code via @privacyscrubber/mcp-server.
How-To: Add 1 JSON block to IDE settings. Whenever the coding agent reads local files, stack traces, or SQL dumps, secrets (.env keys, JWTs, AWS tokens) and customer PII are tokenized in RAM before entering the LLM context window.
Protects Node.js/WASM microservices, ETL jobs, RAG vector DBs & CI/CD via @privacyscrubber/sdk.
How-To: Call scrubText(payload, { profile: 'medical' }) before storing embeddings in Pinecone / Weaviate / pgvector or calling OpenAI/Anthropic APIs, and unscrubText() on return — executing in <0.4ms with 0 network egress.
Security tools fail when they slow teams down. In the Teams Governance Hub ($99/mo flat for unlimited seats), PrivacyScrubber accelerates collaboration across departments through Company-Wide Auto-Scrub and Zero-Server Session Handoff.
| Workflow Step | What Happens on Screen (Simple How-To) | Why the CISO Trusts It |
|---|---|---|
| Step 1: Analyst Scrubs & Prompts AI | An analyst drops a confidential M&A contract or incident log, sends the tokenized prompt to ChatGPT/Claude, and gets the AI's strategic output containing tokens ([NAME_1], [ORG_1]). |
The external AI vendor never sees a single real name, account number, or secret. |
| Step 2: 1-Click Encrypted Handoff | Instead of pasting raw secrets into Slack or Email, the analyst clicks Share Session and sends the sealed handoff package to a Manager, Partner, or teammate. | Protected by Argon2id memory-hard key derivation + XChaCha20-Poly1305 (256-bit authenticated encryption). Zero keys or data touch any server. |
| Step 3: Instant Local Reveal by Colleague | The Manager opens the handoff in their browser or extension and clicks Reveal — instantly seeing the full AI analysis with all real client names and figures restored. | 100% peer-to-peer decryption in the recipient's local RAM. Slack, Teams, Email, and the AI cloud hold zero cleartext PII. |
One flat subscription gives every role in your company the exact controls they need — combining un-bypassable CISO governance with zero-ticket employee speed:
×, and reveal replies in 1 click.| Regulatory Standard | Specific Clause | PrivacyScrubber ZTDS™ Control | Auditor Evidence |
|---|---|---|---|
| EU AI Act (Enforced Aug 2026) | Article 50 & 53 (Data Minimization & Transparency) | Pre-AI client-side tokenization ensures no employee or customer PII enters foundation model training pipelines. Shifts compliance left to the endpoint. | Data Protection Impact Assessment (DPIA) + 1-click PDF Audit Receipts confirming zero cloud egress. |
| Composable Privacy Stack | Gartner Shift-Left Architecture | Zero-server modular privacy layer. Drops into browser, IDE & backend workflows with zero infrastructure overhead and zero vendor lock-in. | Zero-server RAM benchmark proof + local cryptographic hash statements. |
| SOC 2 Type II | CC9.1 (Confidentiality & Data Protection) | In-memory ZTDS™ isolation prevents unauthorized third-party disclosure of customer data to LLMs. | Network Inspection Log showing 0 outbound requests + ZTDS™ System Description. |
| ISO/IEC 27001:2022 | Control A.8.11 (Data Masking) | Mandatory client-side pseudonymization before sending text to external cloud APIs. | Airplane Mode Audit Screenshots + ISMS Policy linking to PrivacyScrubber endpoint. |
| NIST AI RMF 1.0 | GOVERN 1.2 & MAP 1.5 | Technical boundary eliminating corporate training data poisoning and prompt leakage. | Third-party vendor risk assessment classifying AI models as "Untrusted Processors". |
| HIPAA | 45 CFR § 164.514(b) (Safe Harbor) | Removes all 18 PHI identifiers prior to AI analysis, permitting AI use without vendor BAAs. | Cryptographic Audit Receipts confirming zero PHI transmission. |
| GDPR & CCPA / CPRA | GDPR Art. 5(1)(c), 25, 28 & 32; CCPA §1798.100 | 100% Direct PII Zero-Egress in client RAM + contextual entropy reduction across 30 vertical profiles. PrivacyScrubber holds Zero Data Processor status (Art. 28). | Statutory Risk Minimization & Technical Safe Harbor (Defense-in-Depth) shielding DPOs and directors from administrative liability. |
You never have to take a vendor's word on security. For SOC 2 Type II walkthroughs, ISO 27001 audits, and board reviews, any CISO or auditor can empirically prove that PrivacyScrubber enforces 100% Zero-Egress ZTDS™ Conformance in 60 seconds:
Open https://privacyscrubber.com in Chrome or Edge. Open DevTools (F12 → Network tab), select Fetch/XHR and WS, and clear the log.
Disconnect Wi-Fi and unplug Ethernet (or switch DevTools to Offline). The workstation is now physically air-gapped from the internet.
Paste a live contract, patient record, or source code snippet (or drop a PDF/Excel file) and click Scrub PII.
Observe instant tokenization and 1-click Reveal working offline. Check the Network tab: confirm exactly zero (0) outbound packets were emitted.
Click Compliance Cert to generate a client-side PDF attestation with the session's SHA-256 signature, risk score, and entity count for your audit binder.
Because 100% of prompt and document sanitization executes inside local endpoint RAM, PrivacyScrubber acts strictly as a local software appliance — never a Data Processor or Subprocessor under GDPR Article 28. Formal mathematical proofs, IETF Internet-Draft specifications (draft-sibiryakov-ztds-protocol), and independent conformance criteria are published on the vendor-neutral consortium portal at https://ztds.ai/standard/, with administrative billing/licensing boundaries detailed in our Privacy Policy.
Legacy security vendors punish enterprise growth with per-seat taxes and token volume overages. PrivacyScrubber replaces $4.88M breach exposure and $150,000/yr DLP contracts with predictable, flat-rate unlimited licensing:
| Cost / Risk Vector | Legacy Per-Seat DLP / Cloud Proxy | PrivacyScrubber Flat-Rate Appliance |
|---|---|---|
| Regulatory Fine & Officer Liability (GDPR / EU AI Act) | Up to €20,000,000 or 4% of global turnover if cloud proxy or AI vendor logs are breached. | Statutory Risk Minimization & Technical Safe Harbor (GDPR Art. 25 & 32 / HIPAA §164.514 / SOC 2 CC9.1): 100% direct PII zero-egress in client RAM + 30 vertical profiles. |
| Corporate IP & Breach Exposure | $4,880,000 average enterprise data breach cost (IBM Report) + 6-month proxy rollout delay. | 60-Second Rollout: Source code, M&A terms, and client PII are tokenized locally before leaving the device. |
| Annual Licensing for 250–1,000+ Staff | $120,000 – $350,000 / year ($30–$50/seat/mo + cloud bandwidth & maintenance fees). | $99/mo Flat (TEAMS) for Unlimited Seats & Unlimited Volume across the entire organization ($0 server overhead). |
[LOCAL] rules & Offline Studio.Argon2id + XChaCha20-Poly1305.@privacyscrubber/mcp-server coverage for Cursor, Claude Code & Windsurf.
• Community Edition ($0 Free): Full client-side sanitization, 15k chars (General Profile), 5k free trial quota across all 30 specialized industry profiles, Developer SDK & MCP Server free with quota.
• PRO Edition ($15/mo or $110 Lifetime): Individual power-user license — all 30 specialized industry profiles with unlimited characters, custom entity & pattern rules, offline PDF/Excel/OCR document studio.
• TEAMS Edition ($99/mo Flat Rate — Unlimited Seats & Unlimited Volume): Full company/department rollout with Admin Hard-Lock, Company-Wide Auto-Scrub, Worker [LOCAL] rules, Encrypted Session Handoff, and signed Audit Receipts.
• DEVELOPER SDK ($299/mo or $2,990/yr) — Unlimited Backend Nodes: Headless Node.js & WASM appliance (@privacyscrubber/sdk & @privacyscrubber/mcp-server) for unlimited internal microservices, CI/CD gates, and RAG vector pipelines (<0.4ms latency, 0 network egress).
• ENTERPRISE Edition (Custom): Air-gapped on-premise distribution, GRC compliance support, and full source code escrow licensing.
In a regulatory investigation or SOC 2 / ISO 27001 / GDPR audit, written policies are no longer sufficient — auditors demand empirical proof of technical enforcement. PrivacyScrubber shifts your organization from policy-based hope to deterministic cryptographic proof. Every sanitization session generates an immutable SHA-256 compliance receipt in local RAM, backed by our independently verifiable ZTDS™ RFC v1.0 Conformance Certificate — establishing Statutory Risk Minimization & Technical Safe Harbor (GDPR Art. 25 & 32, HIPAA §164.514(b), SOC 2 CC9.1, EU AI Act Art. 50/53) and shielding CISO/DPO leadership from personal administrative liability.
ZTDS-CERT-2026-PRIVACYSCRUBBER-WEB-D02DC6
Your company does not need to spend 6 months or $150,000+ building custom sanitization infrastructure to achieve full ZTDS™ RFC v1.0 compliance and earn the official Corporate Adopter Safe Harbor Seal on your side. Follow the optimal 2-step enterprise pathway: