A self-hosted governance gateway for Southeast Asian enterprises. RedactNode detects Malaysian PII locally, tokenizes identifiers before dispatch to public LLMs, and records immutable SHA-256 audit trails.
Tanda Guard Chrome MV3 extension intercepts unmasked prompts directly on web chat interfaces.
FastAPI proxy masks Malaysian MyKad, phone numbers, and secrets before external transit.
Safe Zone Next.js portal provides protected document chat, OCR redaction, and audited reveals.
Test a sample enterprise prompt containing Malaysian identifiers or secrets. Toggle between a protected Tanda Gateway transit and an unmanaged direct submission.
Raw PII remains inside your self-hosted perimeter. Third-party providers receive only contextual tokens, allowing safe reasoning without permanent identifier exposure.
Enforces strict consent boundaries and cross-border transfer controls before prompt data reaches external providers.
Aligns with Bank Negara Malaysia risk management rules for third-party technology and cloud integrations.
Guarantees patient identification numbers (MyKad) and clinical notes stay strictly inside approved perimeters.
Immutable SHA-256 verification records for internal compliance officer review and direct SIEM webhook export.
RedactNode does not rely on employee policy compliance or manual reminders. It enforces zero-leak privacy across the browser, application gateway, and governance portal.
A Chrome Manifest V3 enterprise extension deployed across staff workstations. Intercepts prompt inputs on ChatGPT, Claude, and Gemini before transmission.
An asynchronous, high-throughput FastAPI gateway running in customer infrastructure. Evaluates prompt payloads before external transit.
An authenticated Next.js portal providing sanctioned AI chat, document extraction, manual PDF redaction, and compliance administration.
RedactNode acts as the verifiable gatekeeper between your workforce and external AI providers. Every request answers two non-negotiable questions: can this sensitive data leave the boundary, and is this model approved?
Real-time evaluation events across customer infrastructure
| Timestamp | Application / Target | Detected Entity | Policy Standard | Action |
|---|---|---|---|---|
| 14:32:11 | ChatGPT Web (Tanda Guard) | MyKad NRIC (850315-14-5521) | PDPA Act 709 | TOKENIZE |
| 14:31:05 | Claude 3.7 (API Gateway) | AWS_ACCESS_KEY_ID | Secrets Defense | BLOCK |
| 14:28:40 | Gemini 2.5 (Tanda Guard) | Public General Inquiry | Baseline Policy | ALLOW |
| 14:20:19 | Safe Zone Portal | Clinical PDF (IC 910214...) | MOH Sovereignty | REDACT |
| 14:15:52 | Internal HR Copilot | Maybank Account (514012...) | BNM RMiT | TOKENIZE |
Client-side checkpoint for public AI interfaces. Inspects inputs locally, blocks paste events, and hands off to Safe Zone.
Internal request boundary. Intercepts prompt payloads, tokenizes Malaysian PII, and terminates secret leaks.
Sanctioned employee workspace with in-memory PDF extraction, permanent manual redaction, and audited reveals.
Immutable SHA-256 ledger for policy violations, retention schedules, and exportable regulatory evidence.
Integrate internal microservices, automated agents, or LangChain / LlamaIndex workflows with zero architectural rewrites. Change the base URL to point at your self-hosted Tanda Proxy.
1from fastapi import FastAPI
2from redactnode.gateway import RedactNodeMiddleware
3
4app = FastAPI()
5
6# Mount the RedactNode Malaysian compliance boundary
7app.add_middleware(
8 RedactNodeMiddleware,
9 ruleset="malaysia_strict", # MyKad, Phone, Banking IDs
10 audit_chain=True, # Immutable SHA-256 hash log
11 token_ttl_seconds=3600, # Transient token lifecycle
12 injection_shield="onnx_v2" # Scan prompt injection & secrets
13)
14
15@app.post("/v1/chat/completions")
16async def chat_proxy(request: Request):
17 # Upstream receives sanitized context:
18 # "Customer [PERSON_1] with NRIC [MYKAD_1] requested loan status."
19 response = await dispatch_to_llm(request.sanitized_payload)
20
21 # Internal authorized caller receives unmasked response
22 return response.restore_tokens()Every sector handles different classes of sensitive data. RedactNode applies specific detection rules and retention thresholds to match your operational boundaries.
Hospital Networks & Private Practice
Enable physicians and administrative staff to summarize diagnostic notes and patient records without transmitting MyKad numbers, mobile contacts, or health identifiers to public LLMs.
Enforced Safeguards:
Commercial Banks, Takaful & Fintech
Provide compliance and underwriting teams with a controlled AI path that masks Malaysian banking account formats, tax identifiers, and credit references under Bank Negara Malaysia RMiT standards.
Enforced Safeguards:
Government-Linked Companies & Corporate Legal
Empower legal, procurement, and commercial teams to analyze confidential tender documents and vendor contracts while preserving internal trade secrets and unannounced commercial terms.
Enforced Safeguards:
Public & Private Universities
Deploy Tanda Guard across university faculties to stop accidental pasting of student identification data and grant researchers a sanctioned portal for working with field datasets.
Enforced Safeguards:
RedactNode is not a multi-tenant public cloud. Every deployment runs within the customer's designated perimeter so raw identifiers never cross unmanaged boundaries.
Single-node testing
Deploy RedactNode on a local machine or staging VPC to benchmark gateway latency, masking fidelity, and audit format.
Included Capabilities:
Private VPC / Cloud
A structured, 30-day proof-of-value deployment for compliance, legal, or IT governance teams in Malaysian organizations.
Included Capabilities:
Dedicated Infrastructure
For GLCs, banks, and healthcare networks requiring air-gapped operations, custom retention, and regulatory sign-off.
Included Capabilities:
Technical analysis of Southeast Asian AI privacy, statutory compliance, and gateway engineering.
A comprehensive guide for Data Protection Officers navigating cross-border transfer rules and AI prompt governance under Act 709.
Why unmanaged employee usage on public LLM interfaces constitutes an uncontrolled egress channel for customer identifiers and confidential memos.
A technical architectural review of Chrome Manifest V3 interception, client-side regex detection, and zero-prompt transit handoffs.
Validate gateway latency, test Malaysian PII detection rules with sample enterprise workloads, and review the hash-chained audit format inside your security boundary.