Oasis Security Warns of NVIDIA NemoClaw Vulnerability Allowing Web-Based AI Model Poisoning
Unauthenticated webpage access could compromise local Ollama instances and inject malicious instructions into AI models
Key Takeaways
- Oasis Security has disclosed a vulnerability in NVIDIA NemoClaw affecting local Ollama instances.
- An attacker-controlled webpage could gain unauthenticated control of the local Ollama instance.
- The flaw may allow planting hidden instructions inside the AI model, enabling prompt injection.
Related Security News
Official MCP Python SDK Flaw Can Let Malicious Servers Steal OAuth Credentials
The maintainers of the official MCP Python SDK disclosed a security vulnerability that could allow a malicious server to trick applications into divulging OAuth credentials. The issue affects the handling of client secrets, authorization codes, and PKCE proof keys when communicating with token endpoints.
Star Blizzard Campaign Targets 100+ Organizations with Fake Event Invitations
Microsoft reports that the Russian state-sponsored threat actor Star Blizzard has been conducting a sustained campaign since January 2026, using fake event invitations to trick targets into installing a backdoor on Windows computers. The operation has affected more than 100 organizations, primarily in the U.S. and U.K., with victims tied to Ukraine. At least one infection has been confirmed, though the full extent of breaches and data exfiltration remains unverified.
