{"id":"CVE-2026-4944","aliases":["EUVD-2026-32979","GHSA-g57c-wgqx-8wx7"],"status":"published","title":"Hardcoded trust_remote_code=True in vllm-project/vllm Bypasses User Security Control","description":"vllm-project/vllm version 0.14.1 contains a vulnerability where the `trust_remote_code=True` parameter is hardcoded in two model implementation files (`vllm/model_executor/models/nemotron_vl.py` and `vllm/model_executor/models/kimi_k25.py`). This bypasses the user's explicit `--trust-remote-code=False` setting, enabling remote code execution via malicious HuggingFace model repositories. This issue is an incomplete fix for CVE-2025-66448 and CVE-2026-22807, as it affects separate code paths in model implementation files. Deployments loading NemotronVL or KimiK25 models are particularly impacted.","severity":{"version":"3.0","score":8.8,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-11","score":0.00747,"percentile":0.553,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":false},"cwes":["CWE-22"],"published":"2026-05-28T18:04:05Z","modified":"2026-06-17T10:57:28Z","products":[{"vendor":"vllm-project","product":"vllm-project/vllm"}],"references":[{"url":"https://huntr.com/bounties/97f706f7-a852-49b2-a4eb-76811e611daf","tags":["exploit"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-4944"},{"url":"https://github.com/advisories/GHSA-g57c-wgqx-8wx7"}],"sources":{"cve":[{"id":"CVE-2026-4944","title":"Hardcoded trust_remote_code=True in vllm-project/vllm Bypasses User Security Control","description":"vllm-project/vllm version 0.14.1 contains a vulnerability where the `trust_remote_code=True` parameter is hardcoded in two model implementation files (`vllm/model_executor/models/nemotron_vl.py` and `vllm/model_executor/models/kimi_k25.py`). This bypasses the user's explicit `--trust-remote-code=False` setting, enabling remote code execution via malicious HuggingFace model repositories. This issue is an incomplete fix for CVE-2025-66448 and CVE-2026-22807, as it affects separate code paths in model implementation files. Deployments loading NemotronVL or KimiK25 models are particularly impacted.","status":"published","published":"2026-05-28T18:04:05.371Z","modified":"2026-05-28T19:47:23.664Z","cvss":[{"version":"3.0","score":8.8,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high","assessor":"cna"}],"cwes":["CWE-22"],"updated_at":"2026-11-02T14:19:59.369956Z"}],"euvd":[{"id":"EUVD-2026-32979","description":"vllm-project/vllm version 0.14.1 contains a vulnerability where the `trust_remote_code=True` parameter is hardcoded in two model implementation files (`vllm/model_executor/models/nemotron_vl.py` and `vllm/model_executor/models/kimi_k25.py`). This bypasses the user's explicit `--trust-remote-code=False` setting, enabling remote code execution via malicious HuggingFace model repositories. This issue is an incomplete fix for CVE-2025-66448 and CVE-2026-22807, as it affects separate code paths in model implementation files. Deployments loading NemotronVL or KimiK25 models are particularly impacted.","status":"published","published":"2026-05-28T18:04:05Z","modified":"2026-05-28T19:47:23Z","cvss":[{"version":"3.0","score":8.8,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high"}],"updated_at":"2026-11-02T18:21:01.680356Z"}],"ghsa":[{"id":"GHSA-g57c-wgqx-8wx7","title":"vllm-project/vllm version 0.14.1 contains a vulnerability where the `trust_remote_code=True`...","description":"vllm-project/vllm version 0.14.1 contains a vulnerability where the `trust_remote_code=True` parameter is hardcoded in two model implementation files (`vllm/model_executor/models/nemotron_vl.py` and `vllm/model_executor/models/kimi_k25.py`). This bypasses the user's explicit `--trust-remote-code=False` setting, enabling remote code execution via malicious HuggingFace model repositories. This issue is an incomplete fix for CVE-2025-66448 and CVE-2026-22807, as it affects separate code paths in model implementation files. Deployments loading NemotronVL or KimiK25 models are particularly impacted.","status":"published","published":"2026-05-28T21:32:03Z","modified":"2026-05-28T21:32:10Z","cvss":[{"version":"3.0","score":8.8,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high"}],"cwes":["CWE-22"],"updated_at":"2026-11-02T15:08:22.24664Z"}],"nvd":[{"id":"CVE-2026-4944","description":"vllm-project/vllm version 0.14.1 contains a vulnerability where the `trust_remote_code=True` parameter is hardcoded in two model implementation files (`vllm/model_executor/models/nemotron_vl.py` and `vllm/model_executor/models/kimi_k25.py`). This bypasses the user's explicit `--trust-remote-code=False` setting, enabling remote code execution via malicious HuggingFace model repositories. This issue is an incomplete fix for CVE-2025-66448 and CVE-2026-22807, as it affects separate code paths in model implementation files. Deployments loading NemotronVL or KimiK25 models are particularly impacted.","status":"published","published":"2026-05-28T19:16:42.677Z","modified":"2026-06-17T10:57:28.91Z","cvss":[{"version":"3.0","score":8.8,"vector":"CVSS:3.0/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","level":"high","assessor":"secondary"}],"cwes":["CWE-22"],"updated_at":"2026-11-02T14:12:21.244426Z"}]}}