Lucene search
K

The vulnerability of the pretrain_gpt() function in the library for training large language models with NVIDIA Megatron-LM allows attackers to execute arbitrary code, gain privileges, and disclose sensitive information.

🗓️ 08 Oct 2025 00:00:00Reported by FSTEC of Russia — Information Security Threat DatabaseType 
bdu_fstec
 bdu_fstec
🔗 bdu.fstec.ru👁 1 Views

The pretrain_gpt vulnerability in Megatron-LM enables code execution, privilege escalation, and data disclosure.

Related
Detection
Refs
ReporterTitlePublishedViews
Family
CNNVD
NVIDIA Megatron-LM 代码注入漏洞
24 Sep 202500:00
cnnvd
CNVD
NVIDIA Megatron-LM Code Injection Vulnerability
28 Sep 202500:00
cnvd
CVE
CVE-2025-23348
24 Sep 202513:13
cve
Cvelist
CVE-2025-23348
24 Sep 202513:13
cvelist
EUVD
EUVD-2025-30969
3 Oct 202520:07
euvd
NVD
CVE-2025-23348
24 Sep 202514:15
nvd
Nvidia
Security Bulletin: NVIDIA Megatron LM - September 2025
23 Sep 202500:00
nvidia
OSV
CVE-2025-23348
24 Sep 202514:15
osv
Positive Technologies
PT-2025-39260
24 Sep 202500:00
ptsecurity
RedhatCVE
CVE-2025-23348
9 Jan 202608:48
redhatcve
Rows per page
Vulners
Node
nvidiamegatron-lmRange<0.13.1
OR
nvidiamegatron-lmRange<0.12.3

Data

Build on a solid foundation with Vulners data

We provide the essential building blocks for cybersecurity solutions with comprehensive, structured, and constantly updated vulnerability and exploits data

Api

Power your application with Vulners API

The Vulners REST API offers reliable, high-performance access to vulnerability intelligence, with 99.9% SLA uptime and CDN-backed data delivery for seamless global access

App

Assess and manage vulnerabilities with Vulners tools

Built on top of Vulners' database and SDK, end-user solutions give security professionals and developers lightweight and powerful tools for vulnerability remediation

08 Oct 2025 00:00Current
6Medium risk
Vulners AI Score6
CVSS 26.8
CVSS 37.8
EPSS0.0022
1