PrivacyGuard FAQs

Question 1

What kind of LLMs does PrivacyGuard integrate with?

Accepted Answer

PrivacyGuard is designed to route sensitive prompts to local Ollama instances, which supports various open-source LLMs such as 'phi3:mini'. This allows you to leverage powerful AI models while maintaining complete control over your data.

Question 2

How do I integrate PrivacyGuard into my existing AI workflow?

Accepted Answer

PrivacyGuard uses the Model Context Protocol (MCP) and exposes its functionalities as tools (`analyze_privacy`, `route_to_local`, `get_audit_logs`) via JSON-RPC over stdio. This enables straightforward configuration and integration with any MCP-compatible client.

Question 3

Can I monitor how PrivacyGuard handles prompts?

Accepted Answer

Yes, PrivacyGuard provides an audit log retrieval feature from its 'audit.log' file. This allows users to review and monitor how sensitive prompts have been classified and routed, enhancing transparency and compliance.

Question 4

How does PrivacyGuard protect my sensitive data?

Accepted Answer

PrivacyGuard automatically classifies prompt text for sensitive keywords. If private information is detected, it routes these specific prompts to a local LLM (like Ollama) within your controlled environment, ensuring the data never leaves your secure infrastructure.

Question 5

What is PrivacyGuard?

Accepted Answer

PrivacyGuard is an infrastructure tool designed to enhance data privacy in AI interactions. It classifies sensitive prompts (containing PII, PHI, or secrets) and routes them to local Large Language Models (LLMs) to prevent confidential information from being exposed externally.

PrivacyGuard

PrivacyGuard

Key Features

Use Cases

Key Features

Use Cases