CVE-2026-122160: myshell-ai AIlice HuggingFace Model Wrapper AModelCausalLM.py LoadModel_PEFT inclusion of functionality from untrusted control sphere
A security vulnerability has been detected in myshell-ai AIlice up to 0.4.0. This vulnerability affects the function LoadModelPEFT of the file ailice/core/llm/AModelCausalLM.py of the component HuggingFace Model Wrapper. The manipulation of the argument basemodelnameorpath leads to inclusion of functionality from untrusted control sphere. It is possible to initiate the attack remotely. The exploit has been disclosed publicly and may be used. The project was informed of the problem early through an issue report but has not responded yet.
Event History
Frequently Asked Questions
Which deployments are affected?
AIlice versions up to and including 0.4.0 are identified as affected, specifically where the HuggingFace Model Wrapper uses LoadModel_PEFT in ailice/core/llm/AModelCausalLM.py.
What does an attacker need to exploit this issue?
The attacker needs to influence the base_model_name_or_path argument passed to LoadModel_PEFT. The issue can be initiated remotely and requires user interaction according to the supplied vector.
How urgent is remediation?
The vulnerability is rated low severity with a 2.1 score, but public exploit information is available and the exploit may be used. No vendor response or fix is reported, so deployments should limit untrusted control over model path inputs until an update is available.