Generative AI guidelines for communications at ETH Zurich
AI is now part of daily comms work. It speeds up ideation, content creation, and review-while raising fair questions about accuracy, privacy, and disclosure.
ETH Zurich has published a practical guide to bring clarity. It's available on Staffnet under "Other Services / Communication."
What the guide covers
- Principles: Protect and strengthen ETH Zurich's credibility. Be transparent. Keep human accountability front and center.
- Data security: Be careful with ETH data. Do not paste sensitive or internal information into public tools. Use approved tools and follow safe handling practices.
- Application and labeling: Clear do's and don'ts for text, images, audio, and video. When and how to state AI involvement. How to attribute AI-generated images and note edits.
- Tool recommendations: Guidance on which tools to consider, and where they make sense in comms workflows.
What this means for PR and Communications
- Editorial use: You can use AI for outlines, drafts, and headlines-but fact-check everything and keep the final call. Label AI involvement where the guide requires it.
- Image creation and edits: Save prompts, keep a record of tools used, and label visuals as AI-generated or AI-edited as specified. Where possible, add provenance or content credentials. For context, see the open standard from C2PA and the industry guidance from Partnership on AI.
- Review and approvals: Treat AI-assisted assets like any external contribution. Run through normal approvals, legal checks when needed, and brand guidelines.
- Privacy: Strip or mask personal or confidential data before using drafting tools. Prefer enterprise or vetted solutions for sensitive tasks.
- Risk management: Watch for outdated facts, bias, and invented sources. Verify names, dates, quotes, and statistics before publishing.
Quick answers to common questions
- Can I publish AI-written text? Yes-after a human edit for accuracy, tone, and policy compliance, and with labeling if required by the guide.
- How do I cite AI-created images? Follow the guide's labeling rules. Include the tool name and an "AI-generated" or "AI-edited" note where applicable.
- Can I put ETH data into AI for brainstorming? Only if the tool and data type are allowed. Do not share sensitive or internal information with public models. Check the data security section of the guide.
Where to find the guide and get support
You'll find "Generative AI in ETH Zurich Communications" on Staffnet under Other Services / Communication. Corporate Communications maintains the guide. Input from the "Netzwerk Kommunikation" and best practices from international academia shaped the content.
Questions about the guide? Use the contact email listed directly below the guidelines on Staffnet.
Translation note
This article has been machine translated for convenience. If something seems off, please refer to the German version. To report a significant translation mistake, email intern-aktuell-news@hk.ethz.ch.
Stay informed
Subscribe to the "internal news" newsletter and check Staffnet regularly to keep up with policy updates and tools that are approved for use.
Want to build skills fast? Explore practical AI courses for communications and marketing roles here: AI courses by job.
Your membership also unlocks:
