CVE-2026-108943
Received Received - Intake

Code Injection in ModelTC LightX2V Tokenizer

Publication date: 2026-10-13

Last updated on: 2026-10-13

Assigner: VulDB

Description
A weakness has been identified in ModelTC LightX2V up to 0.5.0. Affected is the function AutoTokenizer.from_pretrained of the file lightx2v/models/runners/cosmos3/cosmos3_runner.py of the component Tokenizer. This manipulation of the argument auto_map causes code injection. The attack is restricted to local execution. The exploit has been made available to the public and could be used for attacks. 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
2026-10-13
Last Modified
2026-10-13
Generated
2026-10-13
AI Q&A
2026-10-13
EPSS Evaluated
N/A
NVD
CVE-2026-108943
EUVD
EUVD-2026-98182

Affected Vendors & Products

Showing 5 associated CPEs
Vendor Product Version / Range
ModelTC LightX2V 0.1
ModelTC LightX2V 0.2
ModelTC LightX2V 0.3
ModelTC LightX2V 0.4
ModelTC LightX2V 0.5.0

Helpful Resources

Exploitability

CWE
CWE Icon
KEV
KEV Icon
CWE ID Description
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.
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.

Attack-Flow Graph

AI Quick Actions

Instant insights powered by AI
Executive Summary

This vulnerability is a code injection flaw in ModelTC LightX2V up to version 0.5.0. It occurs in the Cosmos3 runner component where the function AutoTokenizer.from_pretrained hard-codes trust_remote_code=True. This allows a malicious model bundle to execute arbitrary Python code during tokenizer loading via the auto_map argument in tokenizer_config.json. The attack is limited to local execution and requires no user authentication.

Detection Guidance

Check for LightX2V versions up to 0.5.0 and inspect tokenizer loading configurations for hard-coded trust_remote_code=True flags. Review model bundles for malicious tokenizer_config.json files with auto_map entries or safetensors metadata containing model_cls values that could redirect execution.

Impact Analysis

If you use LightX2V for image or video generation tasks, an attacker could exploit this to run malicious code on your system with the same privileges as the LightX2V process. This could lead to data theft, system compromise, or further network attacks. The vulnerability affects all versions up to 0.5.0 and requires no user interaction beyond normal model loading.

Compliance Impact

This vulnerability could violate GDPR by enabling unauthorized code execution that may access or exfiltrate personal data. For HIPAA, it risks exposing protected health information if the system processes medical data. Organizations using LightX2V may fail compliance audits due to inadequate code execution controls and lack of input validation.

Mitigation Strategies

Upgrade to a patched version of LightX2V if available. Set trust_remote_code=False in tokenizer configurations. Validate all safetensors metadata against strict allowlists. Audit and sanitize model bundles before loading. Avoid using untrusted model bundles.

Chat Assistant

Ask questions about this CVE
Hi! I’m here to help you understand CVE-2026-108943. Ask me anything about the vulnerability, its impact, or mitigation strategies.
0/70

EPSS Chart