CVE Watchtower

CVE-2026-129984autogpt-platform-beta >= 0.6.67, Missing Authorization

NVD

Vulnerability Summary

A member whose only role in an organization or team was billing manager was treated as an ordinary member by the workflow, run, library and marketplace routes. On a team they could read every other member's runs, and in the organization they could create workflows and marketplace listings. Upgrade to v0.11.3 or later. Impact A team billing manager could read the team's workflows and every member's runs in it, including run inputs, outputs and per-node data, and an organization billing manager could see the organization's shared workflows and runs and create workflows and marketplace listings there. Neither could change or delete other members' workflows, and only an organization or team admin could grant the role, to someone already in the organization. Fix Fixed in AutoGPT Platform v0.11.3. The workflow, run, library, schedule and marketplace routes, External API v1 and the MCP tools now check the caller's role. A role that covers only billing is refused there with a 403 that says why, and keeps its billing access. Fix evidence: 36b8a54561e7. Upgrade to v0.11.3 or later. Version numbers in this advisory identify Platform releases, whose Git tags are `autogpt-platform-beta-v<VERSION>`, rather than the backend Python package version. The `significantgravitas/autogpt` Docker image is tagged `v<VERSION>`. Workarounds Give the billing-manager role only to people who may also see and use the organization's or team's workflows and runs. --- Details Organizations and their teams have a billing-manager role meant for spend. The role matrix in `autogpt_platform/autogpt_libs/autogpt_libs/auth/permissions.py` withheld run access and workflow creation from it, but no workflow, run, library or marketplace route enforced that matrix, and the membership lookup behind those routes ignored the role. - `_TEAM_PERMISSIONS` gave `VIEW_EXECUTIONS`, `CREATE_AGENTS`, `USE_CREDENTIALS` and `DELETE_AGENTS` only to team admins and plain team members. `_ORG_PERMISSIONS` withheld `CREATE_RESOURCES`, `SHARE_RESOURCES` and `PUBLISH_TO_STORE` from billing managers, who "only get finance-related actions". - `requires_team_permission` in `autogpt_platform/autogpt_libs/autogpt_libs/auth/dependencies.py` had no caller, and `requires_org_permission` guarded only organization management, billing and transfer routes. The workflow, run and marketplace routes in `autogpt_platform/backend/backend/api/features/graphs/routes.py`, `graph_executions/routes.py` and `store/routes.py` depended on `get_request_context`, which proved membership and checked no role. - `get_user_team_ids` in `autogpt_platform/backend/backend/data/tenancy.py` returned every active team membership without reading `isBillingManager`, and `visibility_filter` then showed those teams' rows. The run reads in `data/execution.py` and the workflow and library reads used it, so `GET /api/executions` and `GET /api/graphs/{graph_id}/executions/{graph_exec_id}` returned other team members' runs to a team billing manager, with or without an `X-Team-Id` header. Run data can hold prompts, fetched documents and tool results. The extra write access was small: an ordinary member could already create and run workflows in the organization's shared space. The gap was introduced in v0.6.67 by #12670 (commit `0b32f690c2`). References - Fix commit 36b8a54561e7, first released in v0.11.3 - Introduced by #12670 Timeline Reported 2026-11-09 · Triaged 2026-11-09 · Fixed 2026-11-15 · Published pending. Credits This issue was initially discovered by Microsoft's Codename MDASH and validated by Dongkwan Kim and HyungSeok Han of Microsoft Security FORGE Lab. The organization-level part was also found by an automated security scan run by Anthropic.
Severity Level
MEDIUM(4.3)
Published Date
Nov 16, 2026
Last Modified
Nov 16, 2026
Exploitation Status
No confirmed exploitation yet
CVE Record Status
Reserved
This CVE ID is referenced in a public advisory, but the CVE Program has not published its official CVE Record yet. Full CVSS/CPE data from NVD will appear here once it does.
Root Weakness (CWE)
The software does not perform an authorization check when an actor attempts to access a resource or perform an action.
CVSS v3.1 Base Metrics — Score 4.3
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N
Attack VectorNetwork
Attack ComplexityLow
Privileges RequiredLow
User InteractionNone
ScopeUnchanged
ConfidentialityLow
IntegrityNone
AvailabilityNone

Affected & Patched Versions

Affected Versions
  • autogpt-platform-beta >= 0.6.67, < 0.11.3
  • significantgravitas/autogpt >= 0.7.1, < 0.11.3
Patched Versions
  • autogpt-platform-beta 0.11.3
  • significantgravitas/autogpt 0.11.3

What to do

Fix it during your normal patch cycle. No exploitation is confirmed yet.

  1. Check whether you run autogpt-platform-beta >= 0.6.67, and which version.
  2. Update to a patched version listed above, then confirm the new version.

Timeline

  • Nov 16, 2026Published
  • Nov 16, 2026Record last modified

Score, severity and exploitation changes are tracked from the day Watchtower started logging them.

Get notified about new autogpt-platform-beta >= 0.6.67, CVEsYou are reading CVE-2026-129984. Get an email when KEV listings, exploits or new CVEs match your stack.
Set up CVE alerts

Data last checked: Nov 18, 2026 20:01 UTC