The Last Line of Defense Before Your Data Reaches Public AI
Tanda Guard hooks directly into the browser DOM. When an employee writes into ChatGPT, Claude, or Gemini, sensitive Malaysian identifiers are intercepted and tokenized locally before network packets ever leave the machine.
Review the underwriting claim for citizen MyKad 850315-14-5521 and notify account holder at 012-348 9912 regarding approval status.
Sensitive identifiers detected. Cleartext network transmission blocked by Chrome MV3 worker:
Protection engineered for local Malaysian syntax
Generic international DLP models fail on Malaysian identifiers. Tanda Guard integrates purpose-built Malaysian grammars, state code verification, and bank-specific account lengths.
MyKad NRIC Architecture
Unlike naive 12-digit regex, Tanda Guard executes structural validation against National Registration Department (JPN) administrative state codes and valid Gregorian calendar dates.
Malaysian Mobile Numbers
+60 1X-XXX XXXX
Active telco prefixes across Maxis, CelcomDigi, U Mobile, and YTL, formatted or raw unspaced digits.
Commercial Bank Accounts
10-16 Digits (Maybank, CIMB, PBB)
Validates Malaysian bank account formats against Luhn and modulus-11 rules across domestic financial institutions.
Cloud Secrets & API Tokens
AKIA..., sk-..., ghp_...
Detects AWS credentials, OpenAI keys, SSH blocks, and GitHub PATs that risk infrastructure takeover.
How Tanda Guard intercepts inside the browser
Operates entirely in userland memory without network proxy dependencies, VPN configuration, or cloud-side latency.
DOM Event Capture
Hooks input and form submission listeners in the browser DOM. Intercepts draft prompt text before browser network sockets dispatch requests to public AI endpoints.
Sandboxed Web Worker
A dedicated, sandboxed Web Worker evaluates deterministic Malaysian regex grammars, state code tables, and checksum algorithms without blocking the UI thread.
Session Token Vault
Detected PII entities are swapped with deterministic surrogate tokens. Cleartext pairs are preserved strictly in temporary session storage on the client workstation.
Sanitized Egress & Proof
The external AI receives only safe surrogate tokens. An anonymized SHA-256 audit digest is transmitted to your central RedactNode Gateway for compliance telemetry.
Silent deployment for enterprise IT teams
Employees do not need to alter their writing habits or open new tabs. IT administrators maintain fleet-wide enforcement via Microsoft Intune or Google Admin console with locked configuration.
{
"ExtensionSettings": {
"fhlmbjkn...": {
"installation_mode": "force_installed",
"update_url": "https://clients2.google.com/service/update2/crx",
"runtime_allowed_hosts": [
"*://chatgpt.com/*",
"*://claude.ai/*",
"*://*.gemini.google.com/*"
]
}
}
}Chromium Manifest V3 Standard
Built to current Chrome and Microsoft Edge MV3 specifications without deprecated persistent background scripts. Zero CPU overhead during routine web browsing.
Zero-Knowledge Local Execution
Prompts are inspected and tokenized in workstation memory. Raw cleartext personal data is never transmitted to RedactNode servers or external logging infrastructure.
Silent Fleet GPO & Intune Deployment
Centrally push and enforce Tanda Guard across 10,000+ managed enterprise endpoints using standard Chromium Enterprise Policy definitions.
Pre-Configured Frontend Selectors
Continuous DOM updates maintaining compatibility with ChatGPT Enterprise, Anthropic Claude Web, Google Gemini, and Microsoft Copilot.
Deploy Tanda Guard across your enterprise fleet
Install the evaluation extension build, test Malaysian PII detection against synthetic workloads, and review the client-side audit stream inside your security perimeter.