Meta has released Llama Guard 4, a new content safety classifier, now available for integration via the Hugging Face Hub. The model is designed to help developers filter harmful or unsafe content in AI-generated outputs. Llama Guard 4 builds on previous versions with improved accuracy for detecting toxic language, hate speech, and other risky content categories.
Available under a permissive license, the model can be easily deployed alongside Llama 4 or other large language models. Hugging Face's integration provides a seamless pipeline for content moderation, enabling real-time filtering of user inputs and model responses. The release emphasizes Meta's commitment to responsible AI development and providing tools to mitigate misuse of generative models.