# Terms of Service

TU Wien provides its employees and students (hereinafter "users") with AI services through the dataLAB, in particular
[Aqueduct](https://datalab.tuwien.ac.at/aiml/aqueduct/) and [Open WebUI](https://datalab.tuwien.ac.at/aiml/openwebui/). All models run on-premises on TU Wien hardware, see
[Data Protection Information](https://datalab.tuwien.ac.at/aiml/privacy/). By using the services, you agree to comply with these terms and to consider
them binding.

## Who may use the services { #who-may-use-the-services }

- **Employees** of TU Wien with their TU Wien account
- **Students** of TU Wien with approved access, see [Getting Access](https://datalab.tuwien.ac.at/aiml/access/#students)
- **Service accounts** for projects and large jobs, on request

## Intended use { #acceptable-use }

1. The services may only be used for work and study purposes in teaching, research, learning and administration at
   TU Wien.
2. Students' performance must not be assessed in a fully automated way.
3. The services must not be used for calculations and decisions relevant to personnel law.
4. In line with good scientific practice, contributions of AI services to scientific work must be explicitly identified.
   See the
   [recommendations for good practice in dealing with generative AI at TU Wien](https://www.tuwien.at/en/studies/teaching-at-tu-wien/digitally-supported-teaching/artificial-intelligence-in-education/task-force-on-artificial-intelligence-in-teaching-at-the-tu-vienna).
5. It is prohibited to use the services for unlawful purposes, to attack or overload them, or to circumvent their
   security measures or limits.

## Legal requirements { #legal-requirements }

Users must check any (legal) restrictions on using or uploading content themselves and undertake to comply with the
applicable law, in particular copyright law, data protection law and the AI Act.

**Copyright:** check in advance whether copyright-protected content may be processed with an AI service, for example
when uploading a document. Publishers and authors can restrict or exclude the use of their content for AI tools. Content
generated by AI is not protected by copyright, as only humans can be authors. However, if output contains recognisable
parts of a protected work, distributing it may infringe copyright.

**Data protection:** personal data of third parties may only be entered if this is absolutely necessary and the entry is
covered by a legal basis (e.g. public interest or an authorisation under the FOG, the Austrian Research Organisation
Act).

**AI Act:** using the services for prohibited AI practices under
[Art. 5 AI Act](https://eur-lex.europa.eu/eli/reg/2024/1689/oj) is prohibited.

## Web search { #web-search }

When you use search in Open WebUI or the web search MCP server, your queries leave TU Wien and are sent to our external
search provider. This feature is intended exclusively for non-personal data, and processing personal data in any form is
prohibited there. Otherwise, these terms apply accordingly.

## Accounts and API keys { #accounts-and-api-keys }

Passing on your access to third parties, whether for a fee or free of charge, is prohibited. Your account and your API
keys are **personal**:

- **Do not share your API keys** with other people, and do not let others use the services through your account.
- **Keep your keys secret.** Do not put them in code repositories, notebooks, screenshots or other places others can see.
- **You are responsible** for all requests made with your keys.
- **If a key may have leaked**, delete it immediately in the Aqueduct UI and create a new one.
- **Need access for a group, a project or an application?** Ask for a service account in the
  [Matrix room](https://matrix.to/#/#llm-service:tuwien.ac.at) instead of sharing a personal key.

## Limits and fair use { #limits-and-fair-use }

Computing capacity is limited and shared by everyone. Usage limits apply per API key and must not be circumvented, for
example by using other people's keys or accounts. If you need to process a lot of data, contact us first, see
[I need more capacity for a large job](https://datalab.tuwien.ac.at/aiml/faq/#i-need-more-capacity-for-a-large-job-what-can-i-do).

## Model availability { #model-availability }

Models are replaced regularly, usually after 3–4 months and at most after about 6. There is no guarantee that a model
stays available, see [How long does a model stay available?](https://datalab.tuwien.ac.at/aiml/faq/#how-long-does-a-model-stay-available)

## Operation and liability { #availability-of-the-service }

The dataLAB operates the services with the utmost care and to the best of its ability. Continuous, uninterrupted
operation is not guaranteed, and TU Wien accepts no liability in this regard. We announce maintenance and known issues
in the [Matrix room](https://matrix.to/#/#llm-service:tuwien.ac.at). Features and services may be changed, limited or
discontinued.

## Measures in case of violations { #suspension-of-access }

If the services are not used in accordance with these terms, or if applicable law is violated, TU Wien reserves the
right to take immediate measures to reduce the risk, e.g. blocking access, deleting API keys or the content in question,
or taking legal action.

## Changes to these terms { #changes-to-these-terms }

These terms may be amended at any time. The latest version is published on this page. We announce important changes in
the [Matrix room](https://matrix.to/#/#llm-service:tuwien.ac.at).

---

Last updated: 5 November 2026