CVE-2026-4372: A critical remote code execution vulnerability exists in all versions of the HuggingFace transformers library prior to version 5.3.0. The vulnerabilit...

Description

A critical remote code execution vulnerability exists in all versions of the HuggingFace transformers library prior to version 5.3.0. The vulnerability allows an attacker to craft a malicious `config.json` file containing the `_attn_implementation_internal` field set to an attacker-controlled HuggingFace Hub repository ID. When a victim loads this model using the standard `AutoModelForCausalLM.from_pretrained()` API, the library downloads and executes arbitrary Python code from the attacker's repository with the victim's full OS privileges. This issue arises due to unfiltered deserialization of configuration attributes, insufficient sanitization of internal fields, and unsandboxed execution of downloaded kernels. The vulnerability bypasses the `trust_remote_code` security mechanism, is invisible to the victim, and exploits the standard documented usage pattern, making it particularly severe. Users are advised to upgrade to version 5.3.0 or later to mitigate this issue.

Metrics

CVSS 3.0

7.8/10

CVSS:3.0/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H

EPSS Probability

0.48%

37.7th percentile

Probability of exploitation in the next 30 days. Learn more

Weakness Enumeration

Affected Software

Vendor	Product	Versions
Huggingface	Transformers	< 5.3.0

References

https://github.com/huggingface/transformers/commit/a7f8e7ff37d87d1a1a0c8cf607971c607741452fPatch
https://huntr.com/bounties/1f693a6e-6836-4b8b-a0bd-ca036fba8884Exploit, Third Party Advisory

Timeline

Published: May 24, 2026
Last Modified: Jun 17, 2026
Status: Analyzed

Frequently Asked Questions

What is CVE-2026-4372?

A critical remote code execution vulnerability exists in all versions of the HuggingFace transformers library prior to version 5.3.0. The vulnerability allows an attacker to craft a malicious `config.json` file containing the `_attn_implementation_internal` field set to an attacker-controlled HuggingFace Hub repository ID. When a victim loads this model using the standard `AutoModelForCausalLM.from_pretrained()` API, the library downloads and executes arbitrary Python code from the attacker's repository with the victim's full OS privileges. This issue arises due to unfiltered deserialization of configuration attributes, insufficient sanitization of internal fields, and unsandboxed execution of downloaded kernels. The vulnerability bypasses the `trust_remote_code` security mechanism, is invisible to the victim, and exploits the standard documented usage pattern, making it particularly severe. Users are advised to upgrade to version 5.3.0 or later to mitigate this issue.

How severe is CVE-2026-4372?

CVE-2026-4372 has a CVSS score of 7.8/10 (HIGH severity). The EPSS model estimates a 0.48% probability of exploitation in the next 30 days.

How do I fix CVE-2026-4372?

Check the vendor references and advisories linked above for patched versions and mitigation guidance. You can also run a Strix scan to test if your systems are affected.