LLM Guard

LLM Guard is an open-source security toolkit for LLM applications, providing input/output scanners that detect and block prompt injection attempts, PII leakage, toxic content, and other risks in text flowing to and from a language model.

It sits between your application and the LLM API, running configurable scanners (anonymization, prompt-injection detection, sensitive-data detection, output relevance checks) before a request reaches the model or a response reaches the user.

Used by teams shipping LLM-powered features to production who need a security/safety layer beyond just trusting the model provider's own guardrails.