Vulnerability in NVIDIA NemoClaw Poses Risk to Local AI Model Control

Oasis Security has identified a vulnerability in NVIDIA NemoClaw, an open-source AI agent stack, that could allow a malicious webpage to gain unauthenticated control of a local Ollama instance. This flaw enables attackers to inject hidden instructions into AI models. While a fix is available for macOS and Linux, Windows and WSL users await an update, underscoring the critical need for prompt security patches in AI infrastructure.