The vulnerability resides in NVIDIA TensorRT‑LLM, where an unchecked return value can result in a null pointer dereference. This flaw is classified as CWE‑690. When exploited, the attacker can cause the library to crash, leading to a denial of service. The effect is limited to the application using TensorRT‑LLM; a crash will make the service unavailable until it is restarted or the vulnerable code is patched.

All installations of NVIDIA TensorRT‑LLM on any supported platform are potentially affected, as the issue is present in the base library without a version qualifier. The advisory does not specify a particular release, so any current or older build may be vulnerable until a patch is released.

The CVSS score of 5.5 indicates moderate severity. The exploit probability is not available and the vulnerability is not listed in the CISA KEV catalog, implying there are no confirmed exploit instances yet. The attack likely requires delivery of malformed input to TensorRT‑LLM, which may be feasible in environments that expose the library to untrusted data or network traffic. Because the flaw leads to a crash rather than code execution, the risk is primarily to availability rather than confidentiality or integrity.