CVE-2026-5817 Docker Model Runner container-to-host code execution via unsandboxed trust_remote_code in Python inference backends

High 8.8 Published Updated

Attackers can execute arbitrary code on the Docker host by triggering inference with a malicious model due to unsandboxed trust_remote_code settings.

Exploited in the wild
Not known
Login needed
Any signed-in user
Attackable
Local or same network only
Victim action needed
Not needed
Fix
None listed

What is CVE-2026-5817?

CVE-2026-5817 is a high-severity code injection vulnerability in Docker Desktop, published on 22 May 2026. It can only be exploited from the same machine or local network, with no action from the victim.

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.

Weakness: CWE-829

Which versions are affected by CVE-2026-5817?

It affects 1 product.

Products include Docker Docker Desktop.

How do I fix CVE-2026-5817?

No fix has been published for CVE-2026-5817 yet. Limit exposure of the affected software until one is available.

Upgrade, and never evaluate templates, expressions or code built from user input. More about Code injection.

Is CVE-2026-5817 being exploited?

There is no known exploitation of CVE-2026-5817 in the wild and no public exploit code.

Public exploit code: 2 proofs of concept on GitHub.

Sources

Also known as GHSA-cgqp-ww2v-6rjh.

Is CVE-2026-5817 in your software?

Upload an SBOM and sbomify checks it against every vulnerability here, every day.

Check your SBOMs for free