Overview
Severity: MEDIUM | Affected: NVIDIA | Category: tool
NVIDIA has announced the release of NeMo Guardrails 2.0, a major update to its open-source toolkit for building trustworthy and secure LLM-powered applications. This new version introduces several enterprise-grade features designed to combat emerging threats. Key among them is a real-time multimodal defense system capable of detecting complex jailbreaks that combine text and images. The framework also introduces a new declarative language for defining complex, programmable safety policies, moving beyond simple topic-based filtering to allow for nuanced conversational control. Furthermore, NeMo Guardrails 2.0 integrates with confidential computing environments, enabling organizations to run sensitive inference workloads within secure hardware enclaves to protect models and data in use. This release is aimed at empowering developers to implement robust, customizable security layers around their generative AI systems, addressing the growing need for production-ready safety controls.