CVE-2026-1839: Arbitrary Code Execution via Unsafe torch.load() in Trainer Checkpoint Loading in huggingface/transformers
Description
A vulnerability in the HuggingFace Transformers library, specifically in the `Trainer` class, allows for arbitrary code execution. The `_load_rng_state()` method in `src/transformers/trainer.py` at line 3059 calls `torch.load()` without the `weights_only=True` parameter. This issue affects all versions of the library supporting `torch>=2.2` when used with PyTorch versions below 2.6, as the `safe_globals()` context manager provides no protection in these versions. An attacker can exploit this vulnerability by supplying a malicious checkpoint file, such as `rng_state.pth`, which can execute arbitrary code when loaded. The issue is resolved in version v5.0.0rc3.
Vector CVSS:3.0/AV:L/AC:H/PR:N/UI:R/S:U/C:H/I:L/A:H · CWE-502
Tracked companies
Mapped from the CVE’s affected-product data, NVD CPE matches, vendor advisories or package feeds. A mapping means the vendor’s software is named, not that its hosted service was exposed.
Status incidents that name it
No tracked status incident names this CVE.
Affected vendors and products
huggingface: huggingface/transformers
References
- CVE record on cve.org (opens in new tab)
- NVD entry (opens in new tab)
- github.com/huggingface/transformers/commit/03c8082ba4594c9b8d6fe190ca9bed0e5f8ca… (opens in new tab)
- huntr.com/bounties/3c77bb97-e493-493d-9a88-c57f5c536485 (opens in new tab)
- CVE records from the CVE Program (cve.org). CVE® is a registered trademark of The MITRE Corporation. (opens in new tab)
- This product uses data from the NVD API but is not endorsed or certified by the NVD. (opens in new tab)
- Known Exploited Vulnerabilities catalog from CISA. (opens in new tab)
- EPSS scores from FIRST (Exploit Prediction Scoring System). (opens in new tab)