SafeQueryAI Documentation

Privacy-first document Q&A with local RAG

View the Project on GitHub JYOshiro/SafeQueryAI

Security & Privacy

This page defines the current security and privacy model for SafeQueryAI.

Privacy Model Summary

SafeQueryAI uses privacy-first architecture:

What Data Is Processed

During an active session, the system processes:

Where Data Is Processed

What Is Stored Temporarily

Data Deletion Lifecycle

Data is removed by either of these events:

  1. Manual clear through session delete endpoint.
  2. Automatic expiry after 60 minutes of inactivity.

Cleanup actions include:

What Is Not Sent Externally

SafeQueryAI is designed to avoid external document exfiltration:

The backend enforces a loopback-only Ollama base URL at startup, preventing non-local Ollama endpoints.

Dependency and Trust Assumptions

SafeQueryAI depends on:

Trust boundaries to communicate to stakeholders:

Known Limitations

Operational Controls (Current)

Reviewer Notes