CVE-2026-5817

The record

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.

Affected packages

Related vulnerabilities

• CVE-2026-114727 Roundcube Webmail before 1.6.20 and 1.7.x before 1.7.5 passes potentially unsafe content… CWE-829
• CVE-2026-0303 A code execution vulnerability in Palo Alto Networks Checkov by Prisma® Cloud can allow… CWE-829
• CVE-2026-79721 Code execution can occur in versions of the MLflow platform running version 0.0.1 or… CWE-829
• CVE-2026-86504 In JetBrains IntelliJ IDEA before 2026.2.2 missing project-trust confirmation before… CWE-829
• CVE-2026-18252 GitLab has remediated an issue in GitLab EE affecting all versions from 18.9 before… CWE-829
• CVE-2026-76139 A flaw was found in acm-operator-bundle. CWE-829
• CVE-2026-22306 Download of code without integrity check, inclusion of functionality from untrusted… CWE-829
• CVE-2026-62680 Orval generates type-safe JavaScript clients in TypeScript from OpenAPI v3 and Swagger v2… CWE-829