The vulnerability is a memory leak, classified as CWE-401, that prevents proper deallocation of memory after it is no longer needed, leading to gradual exhaustion of system resources. An attacker who can trigger the leak may cause a denial of service on the NVIDIA Triton Inference Server, disrupting inference workloads and potentially affecting availability for dependent services.

All installations of NVIDIA Triton Inference Server running on Linux are affected. No specific version range is listed, so all current and past releases lacking the fix should be considered vulnerable until an update is applied.

The CVSS score of 7.5 indicates a high severity for loss of availability. The EPSS score is less than 1%, and the vulnerability is not listed in the CISA KEV catalog. The attacker would need to send sufficient inference requests or otherwise invoke the memory‑allocation code path to trigger the leak, which is inferred to be remote via the server’s API, although an explicit attack vector is not given.