CVE-2026-24146: 
Triton Inference Server vulnerability analysis and mitigation

CVE-2026-24146 is a denial-of-service vulnerability in NVIDIA Triton Inference Server caused by insufficient input validation when handling requests with a large number of outputs, which can trigger a server crash. It affects all versions of NVIDIA Triton Inference Server prior to r26.02. The vulnerability was published on April 7, 2026, with NVD initial analysis completed on April 16, 2026. It carries a CVSS v3.1 base score of 7.5 (High), assigned by NVIDIA Corporation (GitHub Advisory, NVIDIA Advisory).

The root cause is classified as CWE-789 (Memory Allocation with Excessive Size Value), where the server allocates memory based on an untrusted, excessively large size value without enforcing expected limits. An unauthenticated remote attacker can send crafted inference requests specifying a large number of outputs, causing the server to attempt excessive memory allocation and ultimately crash. No authentication, user interaction, or elevated privileges are required, and the attack can be conducted entirely over the network with low complexity (GitHub Advisory, NVIDIA Advisory).

Successful exploitation results in a crash of the NVIDIA Triton Inference Server, causing a complete loss of availability for inference operations and any dependent applications or pipelines relying on the server. There is no impact to confidentiality or data integrity. The scope is limited to the affected server instance, but in production AI/ML environments, a crash could disrupt critical inference workloads and downstream services (GitHub Advisory).

NVIDIA has released a patch in Triton Inference Server version r26.02; all versions prior to r26.02 should be upgraded immediately (NVIDIA Advisory). As an interim measure, organizations should implement network access controls to restrict exposure of the Triton Inference Server to trusted sources only, and validate or sanitize input data provided to the inference server. Monitoring vendor advisories at the NVIDIA Product Security page is recommended for any additional guidance.

Coverage of this vulnerability was noted across security news outlets including SecurityOnline and Heise (which reported on NVIDIA's security updates for both DALI and Triton Inference Server), as well as aggregators such as VulDB and The Hacker Wire (Heise News). Community reaction has been relatively muted given the absence of a public exploit and the limited scope of impact (availability only). No notable researcher commentary or significant social media discussion has been identified beyond routine vulnerability tracking.