Meta Prompt Guard 2
Lightweight detectors for prompt injection and jailbreak attempts.
Sorting
Meta Llama Guard 3
LLM-based safety classifier for prompts and responses; multiple sizes.
NVIDIA NeMo Guardrails
Programmable guardrails (topic control, PII, jailbreak prevention) for LLM apps.
Google ShieldGemma
Open-weight safety classifiers for inputs/outputs; tune to your policies.
Amazon Bedrock Guardrails
Centralized, reusable guardrails for prompts and responses across models.
Azure AI Content Safety
Configurable content filters for harmful text and images with studio & APIs.
OpenAI Moderation API (omni-moderation-latest)
Multimodal moderation for text & images; granular safety categories and flags.