# The hidden risks of Open WebUI sit in a Docker volume and its license

> Every client chat lives in a single Docker volume, and putting the client's logo on the interface can breach the license.

Original: https://fdetimes.net/en/tools/open-webui-client-deployment-volume-license/

The Open WebUI Quick Start page contains one short passage every FDE should copy into a runbook. Chats, users and configuration all live in the Docker volume, that volume survives updates, and the system should never be run without it.

The second thing to remember is in the license. The Open WebUI license, in force since v0.6.6 on April 19, 2025, prohibits altering, removing or obscuring the "Open WebUI" branding.

Those two passages decide whether a client deployment goes smoothly. Installation is light work. The hard part is who holds the data, and what the client is allowed to do with the interface it is using.

## What does Open WebUI solve at a client site?

The official documentation describes Open WebUI as an extensible, feature-rich, user-friendly self-hosted AI platform. It supports Ollama and OpenAI-compatible APIs, so it is not tied to any provider: the model can run locally on the client's GPU machines or in the cloud.

That makes it a good fit for a situation FDEs meet constantly: the client already has a model or an internal gateway and lacks only a chat interface staff can use straight away. There is no frontend to write. But every connection has to be wired up by hand, because Open WebUI ships with no model of its own.

## Whoever creates the first account controls the system

The Quick Start states plainly that the first account created becomes the administrator, managing users and every setting across the system. On a laptop, that detail is harmless. At a client site, signing up with a personal email to test quickly hands the highest privileges in the system to someone who is about to leave the project.

At the first setup session, follow this order. Mount the volume from the very first run, then let the client's owner create the first account, or use a service account the client owns.

Next, the admin clicks the avatar, opens Settings > Connections in the admin area and enters the client's Ollama endpoint or OpenAI-compatible gateway. If the model picker is still empty, resist fixing the interface first: call the endpoint with `curl` from the machine running the container to find out whether the fault lies in the network, the gateway or the connection credentials.

## The volume is client data, so treat it as a database

Because chats, users and configuration all live in the volume, it is in effect a database of everything staff have asked the model, possibly including sensitive material. If someone deletes it while tidying up a server, every account and all history go with it.

**Key point:** A chat interface can be replaced in an afternoon; a client's chat data, once lost, cannot be recovered.

Before handover, agree three things in writing: which disk the volume sits on and whose infrastructure that is, who is responsible for backups, and how the data is handled when the project ends.

## The number 50 is not a seat limit

The license has one formal exception: rebranding is allowed if the system has 50 or fewer users within a 30-day period. Aliteq, an independent analyst, stresses that this figure is not a cap on seats. It limits only rebranding.

A hypothetical example shows the consequence. A client runs a pilot for 40 people in its legal department and puts the company logo on the interface: still within the exception. Three months later, the client opens it to 300 employees.

The system may keep running, but once users in a 30-day window exceed 50, the client's logo has to come down, unless the client holds an enterprise license that explicitly permits rebranding.

A small logo does not get a pass. Aliteq cites Open WebUI's documentation to the effect that even a 20-pixel-high logo counts as co-branding. So raise the logo question at kickoff rather than leaving it to the end of the rollout plan.

## The enterprise edition: what to ask, and where to get it confirmed

The vendor's homepage lists enterprise features for heavily regulated industries: SSO, RBAC, audit logs, data residency, and on-premises or air-gapped deployment. That is a marketing list, and it does not yet commit the vendor to anything for the version about to be installed.

So take the client's security checklist and mark the lines that depend on those six items, plus the right to rebrand. Email the vendor asking whether each of those items is included in the license the client plans to buy, and ask for a written answer.

Attach that answer to the project file. Any line the vendor has not confirmed goes down as an open risk, not into the plan as an existing feature.

## What to learn first, and how to put it on a CV

The first skill to practice is Docker volumes and networking: where the data lives, how it is backed up, and which route the container takes to reach the model server. The second gets less attention: reading open-source licenses closely enough to know what can and cannot be promised to a client.

On a CV, avoid a vague "built a chatbot". Write that you deployed a self-hosted chat interface connected to internal models, handed admin rights to the client, set up a backup process and checked the license before rollout. If the job posting mentions on-prem or air-gapped, put that line first.

Getting the container running takes a few minutes. Where the volume lives, who holds admin rights and which logo appears on screen are the three questions to answer before the client's first employee logs in.

**Try this week:**

- Run Open WebUI with a mounted volume, create a few chats, delete the container and run it again to see the data survive
- Go to Settings > Connections, add one Ollama connection and one OpenAI-compatible connection, and write the steps up as a short runbook page
- Draft an email to the vendor asking whether the enterprise edition includes SSO, RBAC, audit logs, data residency, air-gapped deployment and the right to rebrand

## Sources

- [Open WebUI](https://docs.openwebui.com/)

- [Quick Start | Open WebUI](https://docs.openwebui.com/getting-started/quick-start/)

- [GitHub - open-webui/open-webui: User-friendly AI Interface (Supports Ollama, OpenAI API, ...)](https://github.com/open-webui/open-webui)

- [Open WebUI License](https://docs.openwebui.com/license/)

- [Open WebUI: Self-Hosted AI Platform](https://openwebui.com/)

- [Open WebUI License, Over 50 Users Without an Enterprise License?](https://aliteq.com/open-webui-license-explained-2026)
