AI Processing & Training Policy

Last updated: January 9, 2026

Our Commitment

Invoice Navigator does NOT use customer data to train AI models.

How We Use AI

PDF Text Extraction

  • Service: Anthropic Claude API
  • Purpose: Extract text from PDF invoices for validation
  • Data sent: PDF content (text only)
  • Data retained by Anthropic: None (transient processing)
  • Training: Explicitly disabled via API configuration

Change Classification (Internal)

  • Service: Anthropic Claude API
  • Purpose: Classify regulatory changes for our automation system
  • Data sent: Public regulatory text only
  • Customer data involved: None

Anthropic's Commitments

Per Anthropic's API Terms of Service:

  • API inputs/outputs are NOT used for model training
  • Data is processed transiently
  • No human review of API content
  • SOC 2 Type II certified

Note: Anthropic's Data Processing Agreement is available on request.

What This Means for You

Your invoice data never trains AI models

Invoice content is not stored by AI providers

Our core validation does NOT use AI (only PDF Converter does)

Audit logs show all AI API calls (Enterprise)

Opt-Out Options

Enterprise customers can:

  • Disable AI-assisted PDF extraction entirely
  • Use traditional OCR-only processing
  • Process on-premises (Enterprise Plus)

Contact enterprise@invoicenavigator.eu to discuss options.

Verification

  • Anthropic DPA available on request
  • Our API configuration explicitly sets training: false
  • Audit logs show all AI API calls (Enterprise)