Governance

PII Redaction

The Spendplane Governance Shield is a pre-egress, pattern- and policy-driven engine that identifies and redacts Personally Identifiable Information (PII) and API secrets before prompts reach an external LLM.

How do I route v0 traffic to Llama-3?

Automated Discovery

The scanner inspects every outbound prompt for emails, phone numbers, IBANs, and API credentials in under 5ms.

Instant Redaction

Sensitive strings are replaced with consistent semantic tokens (e.g., [EMAIL_1], [API_KEY_1]).

Zero Prompt Body Storage

Only redacted entity counts and routing metadata are logged; raw prompt payloads are never persisted.

Pre-Egress Policy Filter

We believe that the best way to handle PII is to keep raw values out of third-party model logs. Spendplane Gateway (or Spendplane AI Safe locally on your device) scrubs sensitive patterns before the prompt reaches the upstream provider.

Pre-Egress Filter (Regex)
# PII Filter Configuration
{
"filters": [
{ "name": "API_KEY", "pattern": "/(?:sk-|ghp_|AKIA)[a-zA-Z0-9_-]{12,}/g", "replacer": "[API_KEY_1]" },
{ "name": "Email", "pattern": "/[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\\.[a-zA-Z]{2,}/g", "replacer": "[EMAIL_1]" }
]
}

Governance Controls & Data Minimization

Governance Shield helps engineering and security teams enforce data minimization before outbound LLM calls. For strict local residency requirements, use Spendplane AI Safe on-device or a self-hosted VPC deployment.

Trace Hub: Manual v1.0.4 / Security Lead Verification: SHIELDED