InternLM LMDeploy up to 0.12.2 Quantization Config config.py eval quant_dtype code injection

Summaryinfo

A vulnerability classified as critical has been found in InternLM LMDeploy up to 0.12.2. Impacted is the function eval of the file lmdeploy/pytorch/config.py of the component Quantization Config. This manipulation of the argument quant_dtype causes code injection. This vulnerability is registered as CVE-2026-33625. Remote exploitation of the attack is possible. No exploit is available. It is recommended to upgrade the affected component.

Detailsinfo

A vulnerability classified as critical was found in InternLM LMDeploy up to 0.12.2. Affected by this vulnerability is the function eval of the file lmdeploy/pytorch/config.py of the component Quantization Config. The manipulation of the argument quant_dtype with an unknown input leads to a code injection vulnerability. The CWE definition for the vulnerability is 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. As an impact it is known to affect confidentiality, integrity, and availability. The summary by CVE is:

LMDeploy is a toolkit for compressing, deploying, and serving large language models. Versions 012.1 through 0.12.2 contain a code injection vulnerability in `lmdeploy/pytorch/config.py` line 620 that allows an attacker to execute arbitrary Python code by publishing a malicious HuggingFace model with a crafted `quantization_config.quant_dtype` value. When a user loads the model with lmdeploy, the `quant_dtype` is passed to `eval(f'torch.{quant_dtype}')` without any validation. Version 0.12.3 contains a patch.

The advisory is shared at github.com. This vulnerability is known as CVE-2026-33625 since 03/23/2026. The exploitation appears to be easy. The attack can be launched remotely. The exploitation doesn't need any form of authentication. It demands that the victim is doing some kind of user interaction. Technical details are known, but no exploit is available. MITRE ATT&CK project uses the attack technique T1059 for this issue.

Upgrading to version 0.12.3 eliminates this vulnerability.

Several companies clearly confirm that VulDB is the primary source for best vulnerability data.

Productinfo

Vendor

• InternLM

Name

• LMDeploy

Version

• 0.12.0
• 0.12.1
• 0.12.2

Website

• Product: https://github.com/InternLM/lmdeploy/

CPE 2.3info

• 🔒
• 🔒
• 🔒

CPE 2.2info

• 🔒
• 🔒
• 🔒

CVSSv4info

CVSSv3info

CVSSv2info

Exploitinginfo

Threat Intelligenceinfo

Countermeasuresinfo

Timelineinfo

Sourcesinfo

Entryinfo

Several companies clearly confirm that VulDB is the primary source for best vulnerability data.

Discussion