Executive brief
PraisonAI Platform, a system for managing multi-agent AI teams, contains a flaw that allows any workspace member to promote themselves or others to the 'owner' role. By exploiting this, a user with the lowest level of access can take full control of a workspace, including the ability to view sensitive data, modify settings, or delete the workspace entirely. This undermines the security boundaries between different users and organizations using the platform.
Technical details
A privilege escalation vulnerability exists in the `POST /workspaces/{workspace_id}/members` endpoint of PraisonAI Platform versions prior to 0.1.4. The endpoint is protected by a `require_workspace_member` dependency that defaults to a `min_role` of 'member', and the underlying `MemberService.add` function fails to perform any caller-permission validation. An authenticated attacker with low-level workspace access can submit a request with an arbitrary `user_id` and set the `role` to 'owner'. This allows the attacker to grant full administrative control over the workspace to a secondary account they control or to inject unauthorized users into other tenants' workspaces. The issue is resolved in version 0.1.4 by enforcing 'owner' role requirements on the endpoint.
Affected products
- MervinPraison praisonai-platform < 0.1.4
Timeline
- 2026-05-19: patched: Fix committed to repository
- 2026-07-21: disclosed: CVE published to NVD