Skip to content
Heist/ feed
Back to feed
HighFix available

MLflow vulnerable to arbitrary file read on unauthenticated servers via prompt model version source

Type: advisorySources: OSV

### Summary On MLflow tracking servers running without authentication, which is the default for mlflow server, a remote attacker can read arbitrary files from the server’s filesystem by creating a model version tagged as a prompt whose source is a schemeless local path. The fix for CVE-2026-2614 in 3.10.0 added a prompt-specific branch to the `CreateModelVersion` handler that rejects `file://` URIs and absolute paths but runs path validation only when the source carries a URL scheme. A schemeless value such as `../../etc`, a plain relative name, `.`, a percent-encoded path, or a Windows drive-letter path is therefore stored unchanged as the model version’s artifact location, and the `model-versions/get-artifact endpoint` later serves files from it with the privileges of the server process. The tag check also keyed on the presence of the `mlflow.prompt.is_prompt` tag rather than its value, so a false value still selected the weaker branch. Servers using `--app-name basic-auth` are also affected: any authenticated user can create a registered model, so a low-privilege account can use this to read the authentication database, including the admin password hash. Affected: MLflow 3.10.0 through 3.16.1. Fixed in 3.17.0 by [mlflow/mlflow#25870](https://github.com/mlflow/mlflow/pull/25870), which accepts only MLflow’s own placeholder sources for prompt versions and requires the prompt tag to be `true`. ### Details When creating a model version with the `mlflow.prompt.is_prompt` tag se

Products
mlflow
Affected versions
3.10.0, 3.10.1, 3.11.0, 3.11.0rc0, 3.11.0rc1, 3.11.1, 3.12.0, 3.12.0rc0, 3.13.0, 3.13.0rc0, 3.14.0, 3.15.0, 3.15.1, 3.15.2, 3.16.0, 3.16.1
Fixed versions
3.17.0
CVEs
CVE-2026-116690
Confirmed fixed CVEs
Not specified
Updated in this feed
2 Nov 2026, 14:56 UTC · Revision 1

Source evidence

MLflow vulnerable to arbitrary file read on unauthenticated servers via prompt model version source

core-osv · Original source

Published 27 Oct 2026, 21:26 UTC

### Summary On MLflow tracking servers running without authentication, which is the default for mlflow server, a remote attacker can read arbitrary files from the server’s filesystem by creating a model version tagged as a prompt whose source is a schemeless local path. The fix for CVE-2026-2614 in 3.10.0 added a prompt-specific branch to the `CreateModelVersion` handler that rejects `file://` URIs and absolute paths but runs path validation only when the source carries a URL scheme. A schemeless value such as `../../etc`, a plain relative name, `.`, a percent-encoded path, or a Windows drive-letter path is therefore stored unchanged as the model version’s artifact location, and the `model-versions/get-artifact endpoint` later serves files from it with the privileges of the server process. The tag check also keyed on the presence of the `mlflow.prompt.is_prompt` tag rather than its value, so a false value still selected the weaker branch. Servers using `--app-name basic-auth` are also affected: any authenticated user can create a registered model, so a low-privilege account can use this to read the authentication database, including the admin password hash. Affected: MLflow 3.10.0 through 3.16.1. Fixed in 3.17.0 by [mlflow/mlflow#25870](https://github.com/mlflow/mlflow/pull/25870), which accepts only MLflow’s own placeholder sources for prompt versions and requires the prompt tag to be `true`. ### Details When creating a model version with the `mlflow.prompt.is_prompt` tag se

osv:GHSA-m65p-9j7q-6w33

mlflow

Fixed versions: 3.17.0

Complete source record
{
  "id": "bffa015a08608ee38f7a0b45848a526981826e3b9a2b4d80950b5d7dfae97593",
  "url": "https://osv.dev/vulnerability/GHSA-m65p-9j7q-6w33",
  "method": "structured",
  "payload": {
    "kind": "advisory",
    "title": "MLflow vulnerable to arbitrary file read on unauthenticated servers via prompt model version source",
    "summary": "### Summary On MLflow tracking servers running without authentication, which is the default for mlflow server, a remote attacker can read arbitrary files from the server’s filesystem by creating a model version tagged as a prompt whose source is a schemeless local path. The fix for CVE-2026-2614 in 3.10.0 added a prompt-specific branch to the `CreateModelVersion` handler that rejects `file://` URIs and absolute paths but runs path validation only when the source carries a URL scheme. A schemeless value such as `../../etc`, a plain relative name, `.`, a percent-encoded path, or a Windows drive-letter path is therefore stored unchanged as the model version’s artifact location, and the `model-versions/get-artifact endpoint` later serves files from it with the privileges of the server process. The tag check also keyed on the presence of the `mlflow.prompt.is_prompt` tag rather than its value, so a false value still selected the weaker branch. Servers using `--app-name basic-auth` are also affected: any authenticated user can create a registered model, so a low-privilege account can use this to read the authentication database, including the admin password hash. Affected: MLflow 3.10.0 through 3.16.1. Fixed in 3.17.0 by [mlflow/mlflow#25870](https://github.com/mlflow/mlflow/pull/25870), which accepts only MLflow’s own placeholder sources for prompt versions and requires the prompt tag to be `true`. ### Details When creating a model version with the `mlflow.prompt.is_prompt` tag se",
    "evidence": "osv:GHSA-m65p-9j7q-6w33",
    "products": [
      {
        "name": {
          "value": "mlflow",
          "evidence": "/affected/0/package/name"
        },
        "prerequisites": [],
        "fixed_versions": [
          {
            "value": "3.17.0",
            "evidence": "/affected/0/ranges"
          }
        ],
        "affected_versions": [
          {
            "value": "3.10.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.10.1",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.11.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.11.0rc0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.11.0rc1",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.11.1",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.12.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.12.0rc0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.13.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.13.0rc0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.14.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.15.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.15.1",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.15.2",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.16.0",
            "evidence": "/affected/0/versions"
          },
          {
            "value": "3.16.1",
            "evidence": "/affected/0/versions"
          }
        ]
      }
    ],
    "severity": "high",
    "structured": {
      "id": "GHSA-m65p-9j7q-6w33",
      "aliases": [
        "CVE-2026-116690",
        "GHSA-m65p-9j7q-6w33"
      ],
      "metrics": [
        {
          "type": "CVSS_V3",
          "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N"
        }
      ],
      "affected": [
        {
          "ranges": [
            {
              "type": "ECOSYSTEM",
              "events": [
                {
                  "introduced": "3.10.0"
                },
                {
                  "fixed": "3.17.0"
                }
              ]
            }
          ],
          "package": {
            "name": "mlflow",
            "purl": "pkg:pypi/mlflow",
            "ecosystem": "PyPI"
          },
          "versions": [
            "3.10.0",
            "3.10.1",
            "3.11.0",
            "3.11.0rc0",
            "3.11.0rc1",
            "3.11.1",
            "3.12.0",
            "3.12.0rc0",
            "3.13.0",
            "3.13.0rc0",
            "3.14.0",
            "3.15.0",
            "3.15.1",
            "3.15.2",
            "3.16.0",
            "3.16.1"
          ],
          "database_specific": {
            "source": "https://github.com/github/advisory-database/blob/main/advisories/github-reviewed/2026/10/GHSA-m65p-9j7q-6w33/GHSA-m65p-9j7q-6w33.json"
          }
        }
      ],
      "modified": "2026-10-27T21:45:04.271041704Z",
      "provider": "osv",
      "references": [
        "https://github.com/mlflow/mlflow/security/advisories/GHSA-m65p-9j7q-6w33",
        "https://github.com/mlflow/mlflow/pull/25870",
        "https://github.com/mlflow/mlflow/commit/81730a2ce4ebcd60d5d18f68c96f422ad7e30772",
        "https://github.com/mlflow/mlflow",
        "https://github.com/mlflow/mlflow/releases/tag/v3.17.0"
      ],
      "provider_metadata": {
        "cwe_ids": [
          "CWE-22"
        ],
        "severity": "HIGH",
        "github_reviewed": true,
        "nvd_published_at": null,
        "github_reviewed_at": "2026-10-27T21:26:56Z"
      }
    },
    "related_cves": [
      {
        "value": "CVE-2026-116690",
        "evidence": "/aliases"
      }
    ],
    "security_relevant": true,
    "severity_evidence": "/database_specific/severity"
  },
  "source_id": "core-osv",
  "document_id": 486112,
  "published_at": "2026-10-27T21:26:56+00:00"
}