LLaMA-Factory through 0.9.5 contains a remote code execution vulnerability that allows attackers with WebUI access to execute arbitrary Python code by supplying a malicious model path in the Chat or Training interfaces. The application passes user-supplied model path input unvalidated into AutoTokenizer.frompretrained() and AutoModel.frompretrained() with a hardcoded trustremotecode=True parameter, causing the Hugging Face transformers library to fetch and execute arbitrary code from a remote or local model repository with the privileges of the server process.
{
"cna_assigner": "VulnCheck",
"osv_generated_from": "https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/58xxx/CVE-2026-58116.json",
"cwe_ids": [
"CWE-829",
"CWE-94"
]
}{
"source": [
"AFFECTED_FIELD",
"DESCRIPTION"
],
"extracted_events": [
{
"introduced": "0"
},
{
"last_affected": "0.9.5"
},
{
"fixed": "0.9.5"
}
]
}