ASR AI Security Radar
Methodology Recent Incidents Get Notified

Back to incidents

AI security incident: LMDeploy vulnerable to arbitrary code execution via eval() of untrusted quant_dtype i...

Source disclosure: September 18, 2026 | Published on Radar: October 14, 2026 | Source: GitHub Security Advisory | Classification confidence: 55%

Source disclosure is when the advisory appeared in this source (or was added to a catalog), not a confirmed date of exploitation. Radar publication is when this page became available.

This incident is part of the public archive and includes explicit AI-related signals from the cited source material. Review methodology.

LMDeploy vulnerable to arbitrary code execution via eval() of untrusted quant dtype in model config loading 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.

Why This Is AI-Related

This page is treated as AI-specific because the source material references copilot, which places the issue inside an AI workflow, model, assistant, or supporting dependency rather than a generic software bulletin.

  • copilot

Affected Workflow

Review AI plugins, copilots, model-serving helpers, CLI tools, and automation runtimes that execute system commands.

Likely Attack Path

An attacker can turn the vulnerable AI-adjacent component into a path for command execution on the host or service runtime.

Impact

The issue can create a path to command execution inside an AI-facing product, plugin, copilot, or supporting service runtime. Severity HIGH. Classification confidence 55%. Source channel GHSA.

Detection And Triage Signals

  • New shell or process activity from AI-facing services
  • Unexpected outbound connections or file writes after prompt or API activity
  • Privilege changes, container escapes, or suspicious job execution logs

Recommended Response

  • Identify every environment that runs the affected AI plugin, assistant, CLI, or supporting package.
  • Patch or isolate the vulnerable component and remove risky execution permissions while validation is in progress.
  • Review process execution, outbound connections, and file-write logs for signs of post-exploitation activity.

Compliance And Business Impact

Code execution paths create immediate risk of host compromise, credential theft, and downstream lateral movement.

Sources

Want alerts like this in real time?

Get notified with incident context, likely impact, and response guidance.

Get Notified

More incidents