Vulnerability Description
vLLM from 0.7.3 before 0.31.0 forwards the request-supplied mm_processor_kwargs.code_revision to Transformers AutoProcessor.from_pretrained, letting clients select which revision of a model's dynamic processor code is imported when trust_remote_code is enabled. An attacker controlling a revision in the served model repository can send a multimodal request to the unauthenticated /tokenize endpoint to execute arbitrary Python in the vLLM API process.
CVSS Score
HIGH
Related Weaknesses (CWE)
References
- https://github.com/vllm-project/vllm/commit/570735520942b6428ae98f1eb442c6d88ada
- https://github.com/vllm-project/vllm/security/advisories/GHSA-h3rc-6mm3-gc2m
- https://www.vulncheck.com/advisories/vllm-0.7.3-before-0.31.0-rce-via-mm-process
FAQ
What is CVE-2026-114053?
CVE-2026-114053 is a vulnerability with a CVSS score of 8.1 (HIGH). vLLM from 0.7.3 before 0.31.0 forwards the request-supplied mm_processor_kwargs.code_revision to Transformers AutoProcessor.from_pretrained, letting clients select which revision of a model's dynamic ...
How severe is CVE-2026-114053?
CVE-2026-114053 has been rated HIGH with a CVSS base score of 8.1/10. Review the CVSS metrics above for detailed severity breakdown.
Is there a patch for CVE-2026-114053?
Check the references section above for vendor advisories and patch information. Review vendor security bulletins for remediation guidance.