Secure Your LLM with PrivyGate
The open-source AI Firewall designed for Southeast Asia. Anonymize PII, comply with PDPA, and keep your customer data within your VPC.
Try It Live
Powered by PrivyGateQuick examples:
1 Your Original Input
2 What LLM Actually Receives (Anonymized)
Your real PII is never sent to the LLM
PII Replacement Mapping
| Type | Original (Your Data) | → | Replacement (Sent to LLM) |
|---|
3 LLM Raw Response (Contains Fake Data)
4 Final Response You Receive (PII Restored)
Fake data automatically replaced back with your original PII
View Full API Response JSON
Processing through Privgate...
Error
Zero Data Leakage
PII like Names, Phones, NRIC, and Credit Cards are replaced with semantic fakes before leaving your network.
SEA Localized
Built-in support for Singapore NRIC, Indonesia NIK, Thai ID, and multi-language NER (English, Chinese, Bahasa).
Reversible & Seamless
Responses are automatically de-anonymized. Your app sees real data; the LLM only sees fake data.
Drop-in Replacement for OpenAI
No need to rewrite your application. Just change the base_url and point it to PrivyGate.
- ✓ Compatible with LangChain & LlamaIndex
- ✓ Supports streaming responses
- ✓ Enterprise Audit Logs (PDPA ready)
from openai import OpenAI
# Point to your PrivyGate instance
client = OpenAI(
base_url="http://privygate:8000/v1",
api_key="any-string"
)
# Send sensitive data safely
response = client.chat.completions.create(
model="gpt-4",
messages=[{
"role": "user",
"content": "My name is Tan Wei Ming, NRIC S1234567Z..."
}]
)
# PrivyGate handles the rest!
Supported Regions & PII Types
Singapore
NRIC, FIN
Indonesia
NIK, NPWP
China
ID, Phone
Thailand
National ID
Malaysia
MyKad
Universal
Email, CC