GCVE Workshop - 22 September 2026 (14:00-18:00), Luxembourg Before The Vulnopticon Conference - Registration

GHSA-MVX4-532P-XFM9

Vulnerability from github – Published: 2026-07-24 16:55 – Updated: 2026-07-24 16:55
VLAI
Summary
Open WebUI: Scheduled automations continue after pending-user deactivation and stored model ACL revocation
Details

Title: Scheduled automations continue after pending-user deactivation and stored model ACL revocation

Summary

Open WebUI documents pending as a zero-access role used for new sign-ups and deactivated users, and normal HTTP routes enforce that with get_verified_user() (which rejects pending), while automation create/update/run routes additionally require the features.automations permission. Two paths missed that lifecycle gate, so a deactivated (pending) account could keep acting through the background automation scheduler:

  1. Scheduler did not re-gate the owner. When a stored automation became due, execute_automation() rehydrated the owner with Users.get_user_by_id(...) and re-entered the chat completion pipeline without re-checking that the owner was still user/admin or still held features.automations. A still-active automation therefore kept running after its owner was deactivated.
  2. Model ACL only enforced for exact role user. check_model_access() applied private-model grants only when user.role == "user", so a pending principal fell through a branch that denies a normal non-owner user.

Net effect: a deactivated account could continue scheduled chat generation through the background worker, consuming the operator's configured model-provider credentials and reaching a stored automation model ID that its current role/ACL state would no longer permit through normal routes.

Impact

A pending/deactivated account continues to execute due scheduled automations after its access has been revoked, consuming the operator's provider credentials, quota and shared capacity, and bypassing the private-model ACL for the automation's stored model ID. Exploitation requires a previously created active automation and a later transition to pending (deactivation or approval rollback), so it is bounded and not interactive. It does not grant unauthenticated access, account takeover, code execution, or cross-user data exfiltration.

Patched

In 0.10.0:

  • execute_automation() aborts and records an error unless the rehydrated owner is still user or admin and (for non-admins) still holds features.automations, so a deactivated or de-permissioned owner's due automation no longer runs.
  • check_model_access() enforces model ACLs for every non-admin role rather than only the exact role user, so a pending or otherwise unrecognised role no longer falls through.

Credits

@rexpository

Severity
3.1 (Low)
CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:L
Show details on source website
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "open-webui"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0.9.0"
            },
            {
              "fixed": "0.10.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-59226"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-285",
      "CWE-862",
      "CWE-863"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-07-24T16:55:30Z",
    "nvd_published_at": "2026-07-09T17:17:04Z",
    "severity": "LOW"
  },
  "details": "**Title:** Scheduled automations continue after pending-user deactivation and stored model ACL revocation\n\n### Summary\n\nOpen WebUI documents `pending` as a zero-access role used for new sign-ups and deactivated users, and normal HTTP routes enforce that with `get_verified_user()` (which rejects `pending`), while automation create/update/run routes additionally require the `features.automations` permission. Two paths missed that lifecycle gate, so a deactivated (`pending`) account could keep acting through the background automation scheduler:\n\n1. **Scheduler did not re-gate the owner.** When a stored automation became due, `execute_automation()` rehydrated the owner with `Users.get_user_by_id(...)` and re-entered the chat completion pipeline without re-checking that the owner was still `user`/`admin` or still held `features.automations`. A still-active automation therefore kept running after its owner was deactivated.\n2. **Model ACL only enforced for exact role `user`.** `check_model_access()` applied private-model grants only when `user.role == \"user\"`, so a `pending` principal fell through a branch that denies a normal non-owner `user`.\n\nNet effect: a deactivated account could continue scheduled chat generation through the background worker, consuming the operator\u0027s configured model-provider credentials and reaching a stored automation model ID that its current role/ACL state would no longer permit through normal routes.\n\n### Impact\n\nA `pending`/deactivated account continues to execute due scheduled automations after its access has been revoked, consuming the operator\u0027s provider credentials, quota and shared capacity, and bypassing the private-model ACL for the automation\u0027s stored model ID. Exploitation requires a previously created active automation and a later transition to `pending` (deactivation or approval rollback), so it is bounded and not interactive. It does not grant unauthenticated access, account takeover, code execution, or cross-user data exfiltration.\n\n### Patched\n\nIn 0.10.0:\n\n- `execute_automation()` aborts and records an error unless the rehydrated owner is still `user` or `admin` and (for non-admins) still holds `features.automations`, so a deactivated or de-permissioned owner\u0027s due automation no longer runs.\n- `check_model_access()` enforces model ACLs for every non-admin role rather than only the exact role `user`, so a `pending` or otherwise unrecognised role no longer falls through.\n\n### Credits\n\n@rexpository",
  "id": "GHSA-mvx4-532p-xfm9",
  "modified": "2026-07-24T16:55:30Z",
  "published": "2026-07-24T16:55:30Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/open-webui/open-webui/security/advisories/GHSA-mvx4-532p-xfm9"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-59226"
    },
    {
      "type": "WEB",
      "url": "https://github.com/open-webui/open-webui/pull/26047"
    },
    {
      "type": "WEB",
      "url": "https://github.com/open-webui/open-webui/commit/920b655f4689e2118de928fbc936f6ebd4fed396"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/open-webui/open-webui"
    },
    {
      "type": "WEB",
      "url": "https://github.com/open-webui/open-webui/releases/tag/v0.10.0"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:L",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Open WebUI: Scheduled automations continue after pending-user deactivation and stored model ACL revocation"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.





Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

Sightings

Author Source Type Date Other

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.


Detection rules are retrieved from Rulezet.


Related by attack behaviour

Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.