news
Graphics: NVIDIA, Huawei, and Mesa
-
Trail of Bits ☛ Uncovering memory corruption in NVIDIA Triton (as a new hire) -The Trail of Bits Blog
Both vulnerabilities affect the Triton Inference Server up to and including version 25.06 and earned CVSS scores of 9.8 and CVE assignments (CVE-2025-23310 and CVE-2025-23311). We disclosed them to NVIDIA, and they have been patched in Triton release 25.07 on August 4, 2025.
-
TechRepublic ☛ NVIDIA Triton Vulnerabilities Could Let Attackers Hijack Hey Hi (AI) Inference Servers
Three NVIDIA vulnerabilities allow unauthorised users to obtain the IPC memory key and use it to craft malicious inference requests.
-
Tom's Hardware ☛ Huawei is making its Ascend Hey Hi (AI) GPU software toolkit open-source to better compete against CUDA
Huawei is planning to open-source its Ascend Hey Hi (AI) GPU toolkit in efforts to accelerate development and compete with Nvidia's CUDA software ecosystem.
-
Collabora ☛ Mesa 25.2 brings new hardware support for Nouveau users
Starting with Mesa 25.2, NVK will now advertise support for Blackwell (RTX 50xx series) and Kepler (most GT and GTX 600 series, most GTX 700 series, and some GTX 800 series) GPUs.