---
id: CVE-2025-6638
aliases:
  - GHSA-59p9-h35m-wg4g
  - PYSEC-2026-1981
title: Hugging Face Transformers is vulnerable to ReDoS through its MarianTokenizer
summary: Hugging Face Transformers is vulnerable to ReDoS through its MarianTokenizer
severity: medium
cvss: 5.3
cvssVector: 'CVSS:3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L'
vendor: transformers
product: transformers
ecosystem: pip
affected:
  - transformers < 4.53.0
patched:
  - transformers 4.53.0
published: '2025-09-12'
updated: '2026-09-10'
sourceUpdated: '2026-09-10T03:50:58.113501288Z'
source: OSV
sourceUrl: 'https://osv.dev/vulnerability/GHSA-59p9-h35m-wg4g'
references:
  - url: 'https://nvd.nist.gov/vuln/detail/CVE-2025-6638'
  - url: >-
      https://github.com/huggingface/transformers/commit/47c34fba5c303576560cb29767efb452ff12b8be
  - url: >-
      https://github.com/huggingface/transformers/commit/d37f7517972f67e3f2194c000ed0f87f064e5099
  - url: 'https://github.com/huggingface/transformers'
  - url: 'https://huntr.com/bounties/6a6c933f-9ce8-4ded-8b3b-2c1444c61f36'
tags:
  - osv
  - pip
epss: 0.00529
epssPercentile: 0.4243
ingestedAt: '2026-07-08T18:25:45.940Z'
---

## Overview

A Regular Expression Denial of Service (ReDoS) vulnerability was discovered in the Hugging Face Transformers library, specifically affecting the MarianTokenizer's `remove_language_code()` method. This vulnerability is present in version 4.52.4 and has been fixed in version 4.53.0. The issue arises from inefficient regex processing, which can be exploited by crafted input strings containing malformed language code patterns, leading to excessive CPU consumption and potential denial of service.

## Affected packages

- `transformers < 4.53.0`

## Remediation

Upgrade to a patched release:

- `transformers 4.53.0`