Nemotron 3.5 Content Safety: Customizable Multimodal Safety for Global Enterprise AI
Source
huggingface.co
Date
Context
NVIDIA's Nemotron 3.5 Content Safety lets an application supply its own moderation policy alongside the material being checked. The model evaluates a prompt, an optional image, and an optional assistant response together, then returns safety classifications. According to the release article, policies can suppress irrelevant categories or introduce organization-specific ones.
This gives teams a way to express rules that depend on context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →, such as allowing technical discussion of terminating a software process. An optional mode also generates an explanation of the classification, giving reviewers material for examining how the model interpreted a rule.
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
