{“lastseen”:””,”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.\\n\\nAny container on the Docker network can trigger this by calling the model-runner.docker.internal API to pull a malicious model and request inference.”,”published”:”2026-05-22T19:24:15.662Z”,”modified”:”2026-05-22T19:24:15.662Z”,”type”:”cve”,”title”:”Docker Model Runner container-to-host code execution via unsandboxed trust_remote_code in Python inference backends”,”source”:”Docker”,”references”:”https:\/\/docs.docker.com\/desktop\/release-notes\/#4680″,”id”:”CVE-2026-5817″,”bulletinFamily”:””,”cwe”:[“CWE-829″],”cvelist”:null,”sourceData”:”Docker Docker Desktop 4.62.0″,”sourceHref”:””,”cvss”:{“score”:8.8,”severity”:”HIGH”,”vector”:”CVSS:4.0\/AV:L\/AC:L\/AT:P\/PR:L\/UI:N\/VC:H\/VI:H\/VA:H\/SC:H\/SI:H\/SA:H”,”version”:”4.0″},”cvss2″:{},”cvss3″:{“version”:””,”vectorString”:””,”baseScore”:0,”baseSeverity”:””,”attackVector”:””,”attackComplexity”:””,”privilegesRequired”:””,”userInteraction”:””,”scope”:””,”confidentialityImpact”:””,”integrityImpact”:””,”availabilityImpact”:””,”cvssV3″:{“version”:””,”vectorString”:””,”baseScore”:0,”baseSeverity”:””,”attackVector”:””,”attackComplexity”:””,”privilegesRequired”:””,”userInteraction”:””,”scope”:””,”confidentialityImpact”:””,”integrityImpact”:””,”availabilityImpact”:””}},”href”:””,”category_name”:”CVE”,”post_link”:””,”product”:”Docker Desktop”,”version”:”4.62.0″,”vendor”:”Docker”,”ai_description”:”Unsandboxed code execution vulnerability in Docker Model Runner”,”ai_severity”:”High”,”ai_vendor”:”Docker”,”ai_product”:”Docker Model Runner”,”ai_version”:”4.62.0″,”ai_score”:8.8}<\/p>\n","protected":false},"excerpt":{"rendered":"
{“lastseen”:””,”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…<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[6,8,41,12,15,13,7,11,5],"class_list":["post-56427","post","type-post","status-publish","format-standard","hentry","category-category_cve","tag-cve","tag-cvss","tag-cvss-88","tag-exploit","tag-high","tag-news","tag-security","tag-tapic","tag-vulnerability"],"yoast_head":"\n