Nemotron 3.5 Content Safety: Customizable Multimodal Safety for Global Enterprise AI

Nemotron 3.5 Content Safety is a unified multimodal and multilingual AI safety model that supports custom enterprise policies with auditable reasoning traces.

• •Evaluates user prompts, images, and assistant responses together in a single inference call to catch safety violations across modalities
• •Covers 12 explicitly trained languages and generalizes to ~140 languages through Gemma 3 base model's multilingual capabilities
• •Accepts custom policy specifications that the model reasons over rather than relying solely on built-in taxonomy
• •THINK mode provides step-by-step reasoning traces for compliance, auditing, and policy iteration in regulated industries
• •Built on Google Gemma 3 4B with LoRA adapter for real-time deployment on 8GB+ VRAM GPUs