Bug 2526799 (CVE-2026-80047)

Summary: CVE-2026-80047 transformers: Hugging Face Transformers: Unauthorized file write via GenerativePreTrainedModel.load_custom_generate()
Product: [Other] Security Response Reporter: OSIDB Bzimport <bzimport>
Component: vulnerabilityAssignee: Product Security <prodsec-ir-bot>
Status: NEW --- QA Contact:
Severity: medium Docs Contact:
Priority: medium    
Version: unspecifiedCC: alinfoot, anpicker, bbrownin, bparees, dfreiber, drow, dschmidt, dtrifiro, ebourniv, hasun, ikhan, jburrell, jfula, jlanda, jowilson, jwong, kshier, nyancey, omaciel, ometelka, ptisnovs, rbobbitt, sbunciak, simaishi, stcannon, syedriko, ttakamiy, vkumar, weaton, xdharmai, yguenane
Target Milestone: ---Keywords: Security
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: ---
Doc Text:
A flaw was found in the Hugging Face Transformers library. A remote attacker could exploit this vulnerability by enticing a user to load a model using the GenerativePreTrainedModel.load_custom_generate() function. This function writes remote Python files to the local disk without requiring explicit user consent, bypassing intended security checks. This can result in unauthorized files persisting on the system and potentially lead to the execution of previously cached malicious code during subsequent trusted model loads.
 Story Points: ---
Clone Of: Environment:
Last Closed: Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description OSIDB Bzimport 2026-09-01 14:00:59 UTC 
A vulnerability in Hugging Face Transformers (versions 4.49.0, <= 5.8.1) allows remote Python files to be written to local disk without user consent when using GenerativePreTrainedModel.load_custom_generate(). The function fetches and caches a remote module file before performing the required trust_remote_code consent check, inverting the security model enforced by other code-loading paths (such as AutoConfig, AutoModel, and AutoTokenizer). As a result, attacker‑controlled Python code from custom_generate/generate.py is copied into the user’s ~/.cache/huggingface/modules directory even if the user declines the trust prompt. Although execution is correctly gated, the file write is not reversible and can persist across sessions. This can lead to persistent, unauthorized files on disk and stale cache collisions where cached attacker code may later be executed during trusted model loads. The issue stems from an unconditional file write in dynamic_module_utils.py prior to any trust verification.