CVE-2026-5817

CWE-829Published: May 22, 2026· Updated: May 22, 2026

8.2

CVSS v3.1

EPSS:0.01%probability of exploitation in 30 daysPercentile:2.6th

Official Description

The vllm-metal inference backend in Docker Model Runner on macOS unconditionally sets trust_remote_code=True when loading model tokenizers, and runs without sandboxing. This causes transformers.AutoTokenizer.from_pretrained() to import and execute arbitrary Python files included in any model pulled from an OCI registry, resulting in arbitrary code execution on the Docker host as the Docker Desktop user when inference is triggered.

Any container on the Docker network can trigger this by calling the model-runner.docker.internal API to pull a malicious model and request inference.

NVD Source

Technical Analysis

CVE-2026-5817 requires local access, meaning attackers must already have a foothold on the target system.

Exploitation requires low privileges, which limits the exposure to scenarios where an attacker has already gained initial access.

A successful exploit results in complete confidentiality breach (data exposure), full integrity compromise (data manipulation), availability disruption (denial of service), with a CVSS base score of 8.2.

The vulnerability has a "Changed" scope, meaning successful exploitation can impact components beyond the vulnerable component itself — such as the host operating system or adjacent services.