CVE-2025-12805

CVE-2025-12805 is a high-severity vulnerability found in the Red Hat OpenShift AI (RHOAI) llama-stack-operator. This flaw allows unauthorized network access to Llama Stack services deployed in other namespaces due to the absence of a NetworkPolicy. An authenticated user in one namespace can exploit this to access another user’s Llama Stack instance, potentially viewing or manipulating sensitive data. With a CVSS score of 8.1 (High), the vulnerability has a low attack complexity and requires low privileges to achieve high confidentiality and integrity impacts. Currently, there is no evidence of active exploitation, nor are public exploit modules available, though it has received some community attention.