Arbitrary Code Execution in Docker Model Runner on macOS
CVE-2026-5817

8.8HIGH

Key Information:

Vendor

Docker

Status
Docker Desktop
Vendor
CVE Published:
22 May 2026

Badges

👾 Exploit Exists🟡 Public PoC

What is CVE-2026-5817?

The vllm-metal inference backend in Docker Model Runner on macOS is vulnerable due to its unconditional setting of trust_remote_code=True when loading model tokenizers. This design flaw allows for the execution of arbitrary Python files from any model obtained via OCI registry. As a consequence, when inference is triggered, it can lead to arbitrary code execution on the host system under the Docker Desktop user account. Additionally, any container on the Docker network is capable of exploiting this vulnerability by invoking the model-runner.docker.internal API to pull an affected model and request inference, further complicating the security landscape.

Affected Version(s)

Docker Desktop MacOS 4.62.0 < 4.68.0

Exploit Proof of Concept (PoC)

PoC code is written by security researchers to demonstrate the vulnerability can be exploited. PoC code is also a key component for weaponization which could lead to ransomware.

CVE-2026-5817-PoC
Created by gouldnicholas

References

https://docs.docker.com/desktop/release-notes/#4680
release-notes

CVSS V4

Score:
8.8
Severity:
HIGH
Confidentiality:
High
Integrity:
High
Availability:
High
Attack Vector:
Local
Attack Complexity:
Low
Attack Required:
Physical
Privileges Required:
Undefined
User Interaction:
None

Timeline

  • 🟡

    Public PoC available

  • 👾

    Exploit known to exist

  • Vulnerability published

  • Vulnerability Reserved

Credit

David Rochester (@davidrxchester)
Nicholas Gould (@gouldnicholas)
.