Sign in
Deutsch English Español Polski
Low

Transformers Deserialization of Untrusted Data vulnerability

  • Disclosed: 10.04.2024
  • Updated: 11.10.2026 13:19

The huggingface/transformers library is vulnerable to arbitrary code execution through deserialization of untrusted data within the load_repo_checkpoint() function of the TFPreTrainedModel() class. Attackers can execute arbitrary code and commands by crafting a malicious serialized payload, exploiting the use of pickle.load() on data from potentially untrusted sources. This vulnerability allows for remote code execution (RCE) by deceiving victims into loading a seemingly harmless checkpoint during a normal training process, thereby enabling attackers to execute arbitrary code on the targeted machine.

Affected software

Click a technology to see all vulnerabilities affecting it.

PlatformPackageVulnerable versionsFixed in
Python (PyPI) transformers < 4.38.0 4.38.0

References

CVSS score 3.4/10
CVSS vector
CVSS:3.0/AV:N/AC:H/PR:N/UI:R/S:C/C:N/I:N/A:L
Weakness (CWE)
CWE-502
Platform
Python (PyPI)
Package
transformers
Disclosed
10.04.2024
Updated
11.10.2026 13:19

Sources and licenses

OSV / GitHub Advisory Database - PYSEC-2026-1978 - PyPA Advisory Database; CC BY 4.0

OSV / GitHub Advisory Database - GHSA-37q5-v5qm-c9v8 - CC BY 4.0

Vulnerability history: transformers

All published vulnerabilities of this package (Python (PyPI)), newest first. Shows whether this is a one-off slip or a package that needs watching.

All (32) in the archive
Total entries
32
First → latest
05.2023 09.2026
Without a fix
9
Exploited in attacks
0

Severity breakdown

  • High 18
  • Medium 12
  • Low 2

Disclosures per year

Severity:
  1. Hugging Face Transformers downloads custom generation code before trust consent
    High · 7.8 CVE-2026-80047 >= 4.49.0, <= 5.8.1 no fix
  2. Transformers save_pretrained path traversal allows arbitrary file writes through chat template names
    High · 7.1 CVE-2026-9856 < 5.10.0 5.10.0
  3. huggingface/transformers: Arbitrary Code Execution During Model Initialization in the LightGlue Model Loading Path
    High · 8.0 CVE-2026-5241 < 5.5.0 5.5.0
  4. HuggingFace transformers vulnerable to remote code execution
    High · 7.8 CVE-2026-4372 < 5.3.0 5.3.0
  5. Hugging Face Transformers GLM4 Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of
    High · 7.8 CVE-2025-14930 <= 4.57.1 no fix
  6. Hugging Face Transformers HuBERT convert_config Code Injection Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hu
    High · 7.8 CVE-2025-14928 <= 4.57.0 no fix
  7. Hugging Face Transformers X-CLIP Checkpoint Conversion Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on af
    High · 7.8 CVE-2025-14929 <= 5.0.0-rc0 no fix
  8. Hugging Face Transformers SEW convert_config Code Injection Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Huggi
    High · 7.8 CVE-2025-14926 <= 4.57.0 no fix
  9. Hugging Face Transformers SEW-D convert_config Code Injection Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hug
    High · 7.8 CVE-2025-14927 <= 4.57.0 no fix
  10. Hugging Face Transformers Perceiver Model Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected instal
    High · 7.8 CVE-2025-14920 <= 4.54.1 no fix
  11. Hugging Face Transformers Transformer-XL Model Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected i
    High · 7.8 CVE-2025-14921 <= 4.54.1 no fix
  12. Hugging Face Transformers megatron_gpt2 Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installa
    High · 7.8 CVE-2025-14924 <= 4.55.0 no fix
  13. A vulnerability in the `preprocess_string()` function of the `transformers.testing_utils` module in huggingface/transformers version v4.48.3 allows for a Regular Expression Denial of Service (ReDoS) a
    High · 7.5 CVE-2025-2099 < 4.49.0 4.49.0
  14. Deserialization of Untrusted Data in Hugging Face Transformers
    High · 8.8 CVE-2024-11394 < 4.48.0 4.48.0
  15. Deserialization of Untrusted Data in Hugging Face Transformers
    High · 8.8 CVE-2024-11393 < 4.48.0 4.48.0
  16. Deserialization of Untrusted Data in Hugging Face Transformers
    High · 7.5 CVE-2024-11392 < 4.48.0 4.48.0
  17. Transformers Deserialization of Untrusted Data vulnerability this entry
    Low · 3.4 CVE-2024-3568 < 4.38.0 4.38.0
  18. Deserialization of Untrusted Data in GitHub repository huggingface/transformers prior to 4.36.
    High · 7.8 CVE-2023-7018 < 4.36.0 4.36.0
  19. Deserialization of Untrusted Data in GitHub repository huggingface/transformers prior to 4.36.
    High · 8.8 CVE-2023-6730 < 4.36.0 4.36.0

Know which vulnerabilities affect your application

With the DockRay integration and software inventory enabled, DockRay checks reported component versions against known vulnerabilities and notifies you about confirmed matches. Coverage depends on the supported technology and available advisory data.

Connect your application
DockRay newsletter

Application security in your inbox

Important vulnerabilities, actionable guidance and practical advice for your applications. Choose your technologies and follow what matters to your project.

  • A daily digest of new vulnerabilities, critical ones included.
  • Notifications about important updates.
  • Practical advice on application security and monitoring.

Leave empty to receive updates for every technology.

Privacy policy

Free to subscribe. Unsubscribe with one click in any email.