CVE-2026-47628: 
Triton Inference Server vulnerability analysis and mitigation

CVE-2026-47628 is a denial-of-service vulnerability in NVIDIA Triton Inference Server for Linux caused by allocation of resources without limits or throttling. It affects all versions of Triton Inference Server up to and including version 26.05. The vulnerability was published on August 18, 2026, with a patch made available by September 1, 2026. It carries a CVSS v3.1 base score of 7.5 (High) (Feedly, NVIDIA Product Security).

The root cause is classified as CWE-770 (Allocation of Resources Without Limits or Throttling), meaning the server fails to impose adequate constraints on resource consumption triggered by incoming requests. An unauthenticated remote attacker can send crafted network requests that cause the server to allocate resources without bound, exhausting system memory, CPU, or other resources. No authentication or user interaction is required, and the attack complexity is low, making it straightforward to trigger over any network path to the server (Feedly, NVIDIA Product Security).

Successful exploitation results in a denial-of-service condition, rendering the Triton Inference Server unavailable to legitimate users and applications. There is no impact to confidentiality or integrity — the vulnerability is limited to availability. In AI/ML production environments where Triton serves inference workloads, an outage could disrupt downstream applications and services dependent on model inference (Feedly).

There is no public proof-of-concept exploit and no evidence of active in-the-wild exploitation as of the time of disclosure (Feedly). The EPSS score is approximately 0.346%, indicating a low probability of exploitation in the near term. The vulnerability is classified as automatable (no user interaction required), which lowers the barrier for opportunistic attacks. It does not appear in the CISA Known Exploited Vulnerabilities catalog (Feedly).

NVIDIA has released a patch addressing this vulnerability, available via the NVIDIA Product Security bulletin 5865 (August 2026); users should upgrade Triton Inference Server to a version beyond 26.05 (NVIDIA Product Security). As interim workarounds, administrators should implement network-level access controls to restrict connections to trusted sources only, configure OS- or container-level resource limits (e.g., cgroup memory/CPU caps) to prevent full system exhaustion, and deploy rate limiting or request throttling in front of the Triton service. Isolating the Triton service in a resource-constrained environment (e.g., Kubernetes resource quotas) is also recommended (Feedly).

The vulnerability received routine coverage from vulnerability tracking services and security news aggregators shortly after disclosure, including mentions on VulDB, SecurityOnline, and Hawk-Eye's weekly threat landscape digest (Feedly). Social media activity was limited to automated CVE notification accounts on Mastodon and Bluesky. No notable researcher commentary or significant community debate has been observed.