LLM Guard

LLM Guard is an open-source security toolkit for LLM applications, maintained by Protect AI. It wraps model calls with input scanners (prompt injection, toxicity, secrets, code, banned topics) and output scanners (toxicity, bias, sensitive data, relevance, factual consistency).

It sanitizes prompts, detects harmful language, and prevents data leakage, installable via pip for production use. Protect AI was acquired by Palo Alto Networks. MIT licensed.