LLM Guard

LLM Guard is an open-source security toolkit for LLM applications, originally built by Laiyer AI and now maintained by Protect AI, which Palo Alto Networks acquired in 2025. It runs a configurable set of scanners over prompts and responses to detect and sanitize risks such as prompt injection, sensitive data exposure, toxic language, and harmful output. The scanners compose into input and output pipelines that sit around a model in production. Its scanners cover common risks named in the OWASP Top 10 for LLM applications, including prompt injection and sensitive-data leakage. Several scanners sanitize in place, redacting detected entities or masking secrets so a flagged request continues safely once the risky content is removed.

Key Capabilities:

• Input scanners for prompt injection and jailbreak detection

• PII detection with redaction and anonymization

• Output scanners for toxicity, bias, and harmful content

• Topic and relevance filtering against allowed subjects

• Composable scanner pipelines for inputs and outputs

• MIT-licensed Python package for self-hosted deployment