Glossary: AI Security and Guardrails

Llama Guard 4

Also written: Llama Guard

Meta's 12B open model that labels prompts and replies by hazard category, for text and images.