Is CVE-2026-103040 exploited in the wild?

No. Junglewise Threat Intelligence has no report of CVE-2026-103040 being exploited in the wild.

How severe is CVE-2026-103040?

Critical severity, with a CVSS score of 9.8, published 29 September 2026.

Which products and versions are affected?

ModelTC LightLLM through 1.2.0.

Is there a fix for CVE-2026-103040?

No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.

How is CVE-2026-103040 attacked?

It is attacked over the network, in a software library.

Am I affected by CVE-2026-103040?

Check whether you run ModelTC LightLLM at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

CVE-2026-103040: LightLLM remote code execution in router profiler service

CVE-2026-103040 · Severity: critical · CVSS 9.8 · Published 2026-09-29

LightLLM is a Python-based framework for serving large language models. When profiling is enabled via the --enable_profiling flag, the router exposes an unauthenticated network service that allows attackers to execute arbitrary code with the privileges of the inference process. An attacker can send specially crafted objects over the network to trigger code execution while the model continues serving requests normally, leaving no trace in logs.

Technologies: ModelTC LightLLM. Vendors: ModelTC.

Quick answers

Is CVE-2026-103040 exploited in the wild?: No. Junglewise Threat Intelligence has no report of CVE-2026-103040 being exploited in the wild.
How severe is CVE-2026-103040?: Critical severity, with a CVSS score of 9.8, published 29 September 2026.
Which products and versions are affected?: ModelTC LightLLM through 1.2.0.
Is there a fix for CVE-2026-103040?: No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.
How is CVE-2026-103040 attacked?: It is attacked over the network, in a software library.
Am I affected by CVE-2026-103040?: Check whether you run ModelTC LightLLM at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

Technical details

The vulnerability is a deserialization of untrusted data (CWE-502) combined with missing authentication (CWE-306) in the RPyC profiler service. The RouterProfilerService binds to 0.0.0.0 with allow_pickle enabled and no authentication, allowing unauthenticated network attackers to send pickled Python objects to the exposed_profiler_cmd method. These objects are deserialized within the model inference process, executing arbitrary gadgets with full process privileges. A fix requires disabling pickle deserialization, restricting arguments to safe strings, and binding to localhost only.

Affected products

ModelTC LightLLM through 1.2.0

Timeline

2026-09-29: disclosed

References

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.