Redact PII before it reaches a public LLM
What the gateway catches, and redacts
Sentinel classifies each request before acting, so sensitive content is removed or stopped while ordinary traffic passes through unchanged.
Personal data
Names, emails, SSNs, and account numbers are detected in the prompt and redacted before the request is forwarded.
Secrets and credentials
API keys, tokens, and connection strings pasted into a prompt are stripped so they never reach a third-party model.
Source and customer data
Code, internal documents, and customer records are classified so sensitive content is caught before it leaves.
A record of every action
Each redaction and block is logged, so the evidence a regulator asks for is a by-product of running the gateway.
Redaction is one job of the gateway
PII redaction runs inside SUPERWISE Sentinel, the AI gateway that inspects every LLM request, tracks cost, and fails closed by default. The same inspection that removes personal data produces the audit record aligned to the EU AI Act, NIST AI RMF, GDPR, and HIPAA.
See the full Sentinel AI gatewayPII redaction questions
What buyers ask about redacting sensitive data before it reaches a model
Keep sensitive data out of every prompt
Start with the free Solo edition, no credit card. Move up when Sentinel protects production.