CVE-2026-65105

CVE-2026-65105 is a disclosed security vulnerability assigned a CVE identifier for tracking purposes in public vulnerability databases. For organizations deploying AI systems, this vulnerability may impact underlying infrastructure, dependencies, or third-party components used in machine learning pipelines and model serving environments. Timely identification and remediation of such vulnerabilities is essential for maintaining secure AI governance frameworks and preventing unauthorized access to sensitive training data or model artifacts.

1 item

ResearchGlobal2026-08-26

CVE-2026-65105 Lets Attackers Poison NemoClaw's AI Through a Browser Tab

Researchers at Cyera disclosed CVE-2026-65105, a DNS rebinding vulnerability in Nvidia's NemoClaw that allows an unauthenticated attacker to access a local Ollama model server from a malicious browser tab. The attack enables persistent injection of hidden instructions into the model's chat template, instructions that survive future sessions and are invisible in standard metadata views. Organizations running NemoClaw or Ollama-backed models in agentic workflows face the highest exposure.

NemoClaw Nvidia Ollama CVE-2026-65105 model poisoning DNS rebinding agentic AI supply chain security