Monitor vulnerabilities like this one.
Sign up free to get alerted when software you use is affected.
7.8
CVE-2026-4372: HuggingFace Transformers Library: Malicious Model Download and Execution
CVE-2026-4372
Summary
A severe vulnerability in the HuggingFace Transformers library allows attackers to download and execute malicious code on a user's system without their knowledge. This can happen when using the library to load a model from the internet. To stay safe, update to the latest version of the library (5.3.0 or later) as soon as possible.
Original title
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 `conf...
Original 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.
nvd CVSS3.0
7.8
Vulnerability type
CWE-1066
Published: 24 May 2026 · Updated: 31 May 2026 · First seen: 26 May 2026