NVIDIA announced Nemotron 3.5 Content Safety, a specialized set of models designed to act as safety filters for enterprise AI systems. The models can screen both text and image inputs and outputs, giving organizations customizable control over what content passes through their AI pipelines.
The system is multilingual and multimodal, handling safety checks across dozens of languages and flagging categories including hate speech, violence, sexual content, and self-harm. Organizations can tune the thresholds per category to match their specific compliance requirements and regional regulations.
Nemotron 3.5 Content Safety is available through NVIDIA NIM microservices and on Hugging Face, making it deployable across cloud and on-premises environments. The release reflects the growing demand for production-grade safety tooling as enterprises move AI agents and chatbots into customer-facing roles.