Glossary: AI Security and Guardrails

Llama Prompt Guard 2

Also written: Prompt Guard

Meta's small open classifiers (86M and 22M parameters) that detect injection and jailbreak attempts.