LMDeploy vulnerable to arbitrary code execution via eval() of untrusted quant_dtype in model config loading — GHSA-3HMM-RH5Q-GWWR

Summary

### Summary lmdeploy <= latest contains 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. ### Details…

Timeline

Published by source

2026-09-18

Risk context

• Reported severity or threat level: high.
• Reported CVSS score: 8.8 out of 10.
• Review exposure for the affected product or family: pip: lmdeploy.

Product

pip: lmdeploy

What to do

General, cautious steps (verify details in the official source):

• Prioritize patching or mitigation immediately (treat as actively risky).
• Identify affected product versions in your inventory and verify whether you are impacted.
• Apply vendor patches/updates or recommended mitigations as soon as available.
• Read the official advisory for exact affected versions and remediation steps.

Official advisory

• Read the official source
• CVE Record
• NVD

Evidence and provenance

Canonical advisory ID

GHSA-3HMM-RH5Q-GWWR

Source

GitHub

Vendor

GitHub

Published by source

2026-09-18

Last source activity

2026-09-18

Additional identifiers

CVE-2026-33625

Curated by

hlukh.ch

Related advisories and connections

These advisories are connected by shared CVE, product, vendor, source, or time signals.