CVE-2026-112080
Received Received - Intake

Code Injection in Llama-Chinese Llama-Chinese

Publication date: 2026-10-19

Last updated on: 2026-10-19

Assigner: VulDB

Description
A vulnerability was found in LlamaChinese Llama-Chinese up to 6fa0fffb0dd82fe3cfaa1449ee54a5806d26ae9b. The impacted element is the function get_class_from_dynamic_module of the file examples/chat_gradio_no_merge.py of the component Adapter Config. The manipulation of the argument base_model_name_or_path results in code injection. The attack may be performed from remote. The exploit has been made public and could be used. This product utilizes a rolling release system for continuous delivery, and as such, version information for affected or updated releases is not disclosed. The project was informed of the problem early through an issue report but has not responded yet.

CVSS Scores

EPSS Scores

Probability:
Percentile:

Meta Information

Published
2026-10-19
Last Modified
2026-10-19
Generated
2026-10-19
AI Q&A
2026-10-19
EPSS Evaluated
N/A
NVD
CVE-2026-112080
EUVD
EUVD-2026-104068

Affected Vendors & Products

Showing 1 associated CPE
Vendor Product Version / Range
LlamaChinese Llama-Chinese 6fa0fffb0dd82fe3cfaa1449ee54a5806d26ae9b

Helpful Resources

Exploitability

CWE
CWE Icon
KEV
KEV Icon
CWE ID Description
CWE-74 The product constructs all or part of a command, data structure, or record using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify how it is parsed or interpreted when it is sent to a downstream component.
CWE-94 The product constructs all or part of a code segment using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify the syntax or behavior of the intended code segment.

Attack-Flow Graph

AI Quick Actions

Instant insights powered by AI
Executive Summary

This vulnerability is a code injection flaw in Llama-Chinese Llama-Chinese up to a specific commit. It occurs in the function get_class_from_dynamic_module of the file examples/chat_gradio_no_merge.py. The issue involves the base_model_name_or_path argument, which can be manipulated to redirect the loading path of the tokenizer and base model. The script uses hardcoded trust_remote_code=True, allowing an attacker to load arbitrary tokenizer files and custom model code from a malicious repository, enabling code execution.

Detection Guidance

Check for unauthorized adapter configurations in the Llama-Chinese repository, particularly in adapter_config.json files. Inspect scripts like chat_gradio_no_merge.py for hardcoded trust_remote_code=True settings. Look for unexpected model or tokenizer paths pointing to external repositories.

Impact Analysis

This vulnerability could allow remote attackers to execute arbitrary code on your system by tricking you into loading a malicious adapter. It may lead to unauthorized model redirection, supply chain attacks, and compromised inference integrity. Attackers could gain control over the model's behavior or steal sensitive data processed by the model.

Compliance Impact

This vulnerability could lead to data breaches, unauthorized access, or data integrity issues, which may violate GDPR and HIPAA compliance. GDPR requires protecting personal data, and HIPAA mandates safeguarding protected health information. A successful exploit could result in unauthorized data exposure or manipulation, leading to regulatory penalties and loss of trust.

Mitigation Strategies

Remove hardcoded trust_remote_code=True in scripts. Validate base_model_name_or_path inputs to prevent path redirection. Use secure loading paths for models and tokenizers. Monitor for unauthorized adapter installations or suspicious repository references.

Chat Assistant

Ask questions about this CVE
Hi! I’m here to help you understand CVE-2026-112080. Ask me anything about the vulnerability, its impact, or mitigation strategies.
0/70

EPSS Chart