CVE-2026-96560 LightLLM through 1.2.0 Unauthenticated Remote Code Execution via NCCL PD RPyC Control Channel
Attackers can execute arbitrary code by sending malicious pickled objects to an unauthenticated RPyC channel when NCCL PD mode is enabled.
- Exploited in the wild
- Not known
- Login needed
- Not needed
- Attackable
- Over the internet
- Victim action needed
- Not needed
- Fix
- None listed
What is CVE-2026-96560?
CVE-2026-96560 is a critical unsafe deserialization vulnerability in ModelTC LightLLM, published on 23 Sep 2026. An attacker can exploit it over the internet without logging in, with no action from the victim.
Official description
LightLLM through 1.2.0 contains a remote code execution vulnerability in the KV-transfer worker when started with --pd_trans_mode nccl, which exposes an unauthenticated RPyC control channel that deserializes attacker-supplied data. Attackers can send malicious pickled objects to the exposed RPyC ThreadedServer to execute arbitrary code with the privileges of the LightLLM service account.
Weakness: CWE-502
Which versions are affected by CVE-2026-96560?
How do I fix CVE-2026-96560?
No fix has been published for CVE-2026-96560 yet. Limit exposure of the affected software until one is available.
Upgrade, and only deserialise data from trusted sources or with an allow-list of types. More about Unsafe deserialization.
Is CVE-2026-96560 being exploited?
There is no known exploitation of CVE-2026-96560 in the wild and no public exploit code.
Sources
Also known as GHSA-849v-g89f-q67r.