vLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.18.0, two model implementation files hardcode trust_remote_code=True when loading sub-components, bypassing the users explicit --trust-remote-code=False security opt-out. This enables remote code execution via malicious model repositories even when the user has explicitly disabled remote code trust. Version 0.18.0 patches the issue.
The product does not use or incorrectly uses a protection mechanism that provides sufficient defense against directed attacks against the product.
| Name | Vendor | Start Version | End Version |
|---|---|---|---|
| Vllm | Vllm | 0.10.1 (including) | 0.18.0 (excluding) |
| Red Hat AI Inference Server 3.2 | RedHat | rhaiis/vllm-cuda-rhel9:1779223654 | * |
| Red Hat AI Inference Server 3.2 | RedHat | rhaiis/vllm-rocm-rhel9:1779223651 | * |
| Red Hat AI Inference Server 3.3 | RedHat | rhaiis/vllm-cuda-rhel9:1775680192 | * |
| Red Hat AI Inference Server 3.3 | RedHat | rhaiis/vllm-rocm-rhel9:1775680262 | * |
| Red Hat AI Inference Server 3.3 | RedHat | rhaiis/model-opt-cuda-rhel9:1775749857 | * |
| Red Hat Enterprise Linux AI 3.3 | RedHat | rhelai3/bootc-aws-cuda-rhel9:1776871984 | * |
| Red Hat Enterprise Linux AI 3.3 | RedHat | rhelai3/bootc-azure-cuda-rhel9:1776871985 | * |
| Red Hat Enterprise Linux AI 3.3 | RedHat | rhelai3/bootc-azure-rocm-rhel9:1776872005 | * |
| Red Hat Enterprise Linux AI 3.3 | RedHat | rhelai3/bootc-cuda-rhel9:1776773390 | * |
| Red Hat Enterprise Linux AI 3.3 | RedHat | rhelai3/bootc-gcp-cuda-rhel9:1776871987 | * |
| Red Hat Enterprise Linux AI 3.3 | RedHat | rhelai3/bootc-rocm-rhel9:1776773505 | * |
| Red Hat Enterprise Linux AI 3.3 | RedHat | rhelai3/disk-image-cuda-rhel9:1776938871 | * |
| Red Hat OpenShift AI 2.25 | RedHat | rhoai/odh-vllm-gaudi-rhel9:1780069069 | * |
| Red Hat OpenShift AI 3.3 | RedHat | rhoai/odh-vllm-gaudi-rhel9:1778600187 | * |