训练框架加载检查点即执行代码,AI 训练管线中的模型来源成为攻击面,CVSS 7.8。
ai_agent反序列化rcecve
NVIDIA Megatron-LM Checkpoint Deserialization of Untrusted Data Remote Code Execution Vulnerability
November 10th, 2026
ZDI-26-904
ZDI-CAN-29291
CVE ID
CVE-2026-65110
CVSS Score
7.8
AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Affected Vendors
NVIDIA
Affected Products
Megatron
Vulnerability Details
This vulnerability allows remote attackers to execute arbitrary code on affected installations of NVIDIA Megatron-LM. User interaction is required to exploit this vulnerability in that the target must open a malicious file.
The specific flaw exists within the parsing of checkpoints. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute code in the context of the current process.
Additional Details
Fixed in Megatron-LM version 0.17.1 (
https://github.com/NVIDIA/product-security/tree/main/2026/5881
)
Disclosure Timeline
2026-06-02
- Vulnerability reported to vendor
2026-11-10
- Coordinated public release of advisory
2026-11-10
- Advisory Updated
Credit
winters0x64
Back to Advisories