CVE-2026-1839: Arbitrary Code Execution via Unsafe torch.load() in Trainer Checkpoint Loading in huggingface/transformers
Published Apr 7, 2026
·Updated
A vulnerability in the HuggingFace Transformers library, specifically in the Trainer class, allows for arbitrary code execution. The loadrngstate() method in src/transformers/trainer.py at line 3059 calls torch.load() without the weightsonly=True parameter. This issue affects all versions of the library supporting torch>=2.2 when used with PyTorch versions below 2.6, as the safeglobals() context manager provides no protection in these versions. An attacker can exploit this vulnerability by supplying a malicious checkpoint file, such as rngstate.pth, which can execute arbitrary code when loaded. The issue is resolved in version v5.0.0rc3.
Affected Software
5 affected componentsFixes available
pip/transformers<5.0.0rc3
5.0.0rc3
huggingface transformers<5.0.0
huggingface transformers=5.0.0-rc0
huggingface transformers=5.0.0-rc1
huggingface transformers=5.0.0-rc2
Remediation
Event History
Apr 7, 2026
CVE Published
via MITRE·05:22 AM
Data Sourced
via MITRE·05:22 AM
DescriptionSeverityWeakness
Data Sourced
via NVD·06:16 AM
DescriptionSeverityWeakness
Data Sourced
via NVD·06:16 AM
RemedyAffected Software
Advisory Published
via GitHub·06:30 AM
Data Sourced
via GitHub·06:30 AM
DescriptionSeverityWeaknessAffected Software