Remote code execution via trusted model loading
Published Jun 30, 2026 · Updated Jun 30, 2026
Code injection in hiyouga LlamaFactory 0.9.5 and earlier allows remote attackers with WebUI access to execute Python code via a model path. The Chat and Training loaders pass the submitted path to Hugging Face AutoTokenizer.from_pretrained() and AutoModel.from_pretrained() while forcing trust_remote_code=True, so repository Python modules execute without validating the source. Reachability requires access to the WebUI; successful code runs with the server process's privileges and can read, alter, or interrupt resources available to that process.
Summary
What happened
Code injection in hiyouga LlamaFactory 0.9.5 and earlier allows remote attackers with WebUI access to execute Python code via a model path. The Chat and Training loaders pass the submitted path to Hugging Face AutoTokenizer.from_pretrained() and AutoModel.from_pretrained() while forcing trust_remote_code=True, so repository Python modules execute without validating the source. Reachability requires access to the WebUI; successful code runs with the server process's privileges and can read, alter, or interrupt resources available to that process.
The record
- CVE
- CVE-2026-58116
- Published
- Jun 30, 2026
- Updated
- Jun 30, 2026
- Vendor
- hiyouga
- Product
- LlamaFactory
- Classifications
- CWE-94, CWE-829, T1190, T1059.006
- Attack vector
- network
- Privileges
- unauthenticated
Timeline
How it unfolded
- Jun 30, 2026CVE publishedPublication date reported by the CVE source.
- Jun 30, 2026Record updatedLatest update available in the CVE record.
Exploitability
Present is not the same as exploitable
Compare your product and version with the public record. A matching version still requires validation against your environment.
Is a vulnerable build present?
Compare these published version ranges with your installed build and any vendor patches.
- Affected versionversion=0 <=0.9.5
What conditions does exploitation require?
What is affected?
Attacks
What attackers are doing with it
Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.
Weakness, pattern, technique
Public exploit references
- LlamaFactory WebUI model-path proof of conceptproof of concept · demonstrated
Labels summarize the accepted research assessment. They do not indicate a test against your environment.
Technologies
Your stack
See the directory against your own environment.
Your stack
Check the software in your environment
Book a demo to see how Hinoki identifies affected software and validates exploitability in your environment.
Book a demo