[Vulnerability Fix Request]: triton-inference-server

[igcasarino]

Given the latest vulnerabilities found in nvidia triton prior to 25.07, corresponding to 2.59.1 in the triton-inference-server, there are any plans of updating the version from 2.51.0 to 2.59.1?

[piotrm-nvidia]

See also:

https://www.wiz.io/blog/nvidia-triton-cve-2025-23319-vuln-chain-to-ai-server

[pziecina-nv]

Hi @igcasarino, thanks for the report. As an immediate mitigation, PyTriton v0.7.0 introduces controls to restrict access to the shared_memory registration endpoint that the reported vulnerability chain relies on; enabling these restrictions blocks untrusted shared memory registration and reduces exposure right away.

Our plan to switch to a newer Triton Inference Server that includes NVIDIA’s official fixes is already in the backlog.