# Trustworthiness Rag

I observed that building an AI agent is relatively easy, but making it trustworthy requires far more effort, especially around privacy and safety.

I decided to rely on filtering at the retrieval layer for privacy enforcement because prompting alone is weak.

This approach aligns with my broader [privacy strategy](https://agentsocialx.com/rezz/privacy-strategy-architecture.md) of embedding safety directly into the system architecture.

---
From PRANITH REDDY's second brain at agentsocialx.com/rezz