# CVE-2026-5817: Docker Model Runner arbitrary code execution in vllm-metal backend

Source: https://junglewise.ai/threats/cve-2026-5817-docker-model-runner-arbitrary-code-execution-in-vllm-metal-backend  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-05-22. Updated: 2026-06-01  
License: CC BY 4.0 with a link back

CVE-2026-5817 · Severity: high · CVSS 8.2 · Published 2026-05-22

Docker Model Runner on macOS contains a security flaw in its AI model processing component. An attacker can use a malicious AI model to execute unauthorized code on the host computer with the privileges of the Docker Desktop user. This could lead to full system compromise, data theft, or persistent access to the user's machine.

Technologies: [Apple macOS](https://junglewise.ai/threats/technologies/macos), [Docker Desktop](https://junglewise.ai/threats/technologies/docker-desktop). Vendors: [Apple](https://junglewise.ai/threats/vendors/apple), [Docker](https://junglewise.ai/threats/vendors/docker).

## Quick answers

**Is CVE-2026-5817 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-5817 being exploited in the wild.

**How severe is CVE-2026-5817?** High severity, with a CVSS score of 8.2, published 22 May 2026.

**Which products and versions are affected?** Docker Docker Desktop 4.62.0 to 4.68.0.

**Is there a fix for CVE-2026-5817?** No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.

**How is CVE-2026-5817 attacked?** It is attacked with local access or over the network, in a network appliance.

**Am I affected by CVE-2026-5817?** Check whether you run Apple macOS or Docker Desktop. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## Technical details

The vllm-metal inference backend in Docker Model Runner (DMR) on macOS unconditionally sets 'trust_remote_code=True' when calling 'transformers.AutoTokenizer.from_pretrained()'. Because the backend runs without sandboxing, it imports and executes arbitrary Python files included in any model pulled from an OCI registry. An attacker can trigger this by sending a request to the 'model-runner.docker.internal' API from any container on the Docker network to pull a malicious model and initiate inference. This results in arbitrary code execution on the Docker host as the Docker Desktop user. The issue is addressed in Docker Desktop version 4.68.0.

## Affected products

- Docker Docker Desktop 4.62.0 to 4.68.0

## Timeline

- 2026-05-22: disclosed
- 2026-05-22: advisory
- 2026-06-01: other: NIST analysis completed

## References

- [https://docs.docker.com/desktop/release-notes/](https://docs.docker.com/desktop/release-notes/)

## Related threats

- [Docker Desktop token logging information disclosure](https://junglewise.ai/threats/cve-2026-123082-docker-desktop-before-version-4-94-0-could-write-the-signed-in)
- [Apple Bluetooth use-after-free vulnerability](https://junglewise.ai/threats/cve-2026-113970-a-use-after-free-issue-was-addressed-with-improved-memory)
- [Apple iOS, iPadOS, macOS, and related platforms rate limit bypass in AirPlay pairing](https://junglewise.ai/threats/cve-2026-113969-a-logic-issue-was-addressed-with-improved-state-management-this)
- [Apple iOS and iPadOS Bluetooth pairing code rate limit bypass](https://junglewise.ai/threats/cve-2026-113968-a-logic-issue-was-addressed-with-improved-state-management-this)
- [Apple iOS/iPadOS Bluetooth pairing bypass](https://junglewise.ai/threats/cve-2026-113965-this-issue-was-addressed-with-improved-checks-this-issue-is)