OSISM – AI Manifesto

Artificial intelligence (AI) is developing at a rapid pace and could bring about positive changes in various areas of life, such as healthcare, research, the economy and public administration. However, the development and use of AI also comes with considerable risks and raises complex questions around copyright, data protection and resource consumption. And clear ethical and legal guidelines to prevent AI from harming individuals and communities are only just beginning to take shape.

At the same time, AI is increasingly finding its way into software and operating systems, becoming ever more pervasive and accessible. In many areas, the line between human-created and AI-generated content is becoming blurred. This fundamentally challenges our existing understanding of authorship, credibility and authenticity, and consequently creates distrust and uncertainty.

We are convinced that we can make a positive impact with AI if we use it thoughtfully, responsibly, purposefully and transparently. That calls for clear principles: concise, universally applicable and easy to understand.

10 core principles

1. Open dialogue and reflection

We respect differing views and encourage an open discussion about opportunities and concerns. For this purpose, we have established communication channels and regular forums for dialogue.

2. Purposeful use

We view AI as a supporting tool, not as a substitute for human work or decision-making. AI suggestions serve as a basis for further iterations, not as a final version. We only use AI where we expect clear added value in the long term and where its use is unlikely to have any foreseeable negative impact, for instance at later stages of the work.

3. Transparency and labelling

We clearly and prominently label all content that has been created, in whole or in part, using AI. Examples of such labels:

  • “Created with AI assistance (model) and reviewed for content and editorial quality.”
  • “AI-assisted: (model), reviewed.”
  • “AI-[model]–[date]-ok-[initials]”

4. Traceability

Where relevant, we document prompts, memory, model type, purpose, responsible persons, date and decision. We may make this information publicly available where appropriate.

5. Editorial responsibility

We take personal responsibility for the tone, style and content of what we publish. This is how we maintain our credibility.

6. Review

We review all AI suggestions before publishing them. Among other things, we correct errors, inaccuracies and problematic wording, and simplify anything that is needlessly complicated.
We welcome any relevant feedback, to which we respond promptly.

7. Neutrality

We also check that AI-generated content is balanced, representative and free from biased assumptions.

8. Data protection and data minimisation

We do not process any sensitive or personal data in external AI systems. Our prompts contain only the bare minimum of information required and are run with only the bare minimum of access rights necessary.

9. Limits of use

Particularly in legal, medical or safety-related contexts, we use AI suggestions solely as a support tool. Final decisions are made by people and, where necessary, subject to expert approval.

10. Model choice and sustainability

We continuously evaluate the development of new or existing AI models. We favour GDPR-compliant, EU-developed and open-source AI models, ideally hosted locally or by trusted third parties with a data processing agreement.



Where possible, we take an AI model’s energy and resource efficiency into account when selecting it.

Our pledge

✓ Transparency:
You can easily tell whether content has been created wholly or partly with AI.
✓ Accountability:
People review and take responsibility for published content.
✓ Data protection:
We do not pass sensitive or personal data on to AI services.
✓ Corrections:
If errors are reported to us, we respond openly and promptly.

Checklist (internal)

✓ Is there clear added value from using AI?
✓ Is any personal or sensitive data involved?
✓ Has generated content been reviewed and edited where necessary?
✓ Is labelling required?
✓ Has the content been checked for neutrality?
✓ Has approval been granted by the relevant person, if required?

Einladung zur Mit‑Unterzeichnung

Wenn Sie unsere Prinzipien zum transparenten Umgang mit KI teilen, können Sie unser Manifest gern mitunterzeichnen oder es als Grundlage für Ihre eigene Richtlinie übernehmen.

Version 1.1 © 2026. Licensed under CC BY-SA 4.0.