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.
{
"cna_assigner": "VulnCheck",
"cwe_ids": [
"CWE-94"
],
"osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/114xxx/CVE-2026-114053.json"
}{
"extracted_events": [
{
"range": [
{
"introduced": "0.7.3"
},
{
"fixed": "0.31.0"
}
],
"source": "AFFECTED_FIELD",
"strategy": "StandardRange"
},
{
"range": [
{
"fixed": "570735520942b6428ae98f1eb442c6d88adae55e"
}
],
"source": "REFERENCES_COMMIT"
}
]
}