AI transparency
On this blog I write about AI systems engineering, and I use the tools I write about myself. This page says exactly where. Not as fine print, but because you should know what you’re reading.
What comes from AI
- Article text – The body text of the posts is generated by a language model. The choice of topic, the thesis, structure, examples and sources come from me.
- Translations – The German original is the reference. All other language versions are translated from it with generative AI and reviewed in a second AI pass.
- Images – Hero images and illustrations are AI-generated. They are deliberately illustrative and show no real people, places or events. Screenshots of real applications are real.
- Podcast episodes – The dialogue script and voices of the audio versions (German only) are fully synthetic. The speakers are not real people. How that pipeline is built is described in AI tools for your own blog.
Who is responsible
Editorial responsibility for all content published here lies with Ruben Vitt. No text goes online unread: I check every post before publication, correct it, cut it, and throw out what isn’t right. Mistakes that still slip through are mine, not the model’s.
If you notice something that is factually wrong or looks made up: write to me. I’ll fix it and note the change in the post.
How content is labeled
Every post carries a notice. It distinguishes two levels:
- Made with AI – Text and images were generated by generative AI. On this blog, that’s the norm.
- AI-assisted – I wrote the text myself; AI was only used for editing or research.
Posts without generative AI carry no notice. The level appears at the top of the post next to the date and reading time, so you see it before you start reading. It links to the full notice, including editorial responsibility, at the end of the post.
Legal framework
Since August 2, 2026, the transparency obligations of Article 50 of the EU AI Act (Regulation (EU) 2024/1689) apply. For published AI-generated text, Article 50(4) requires disclosure where the text informs the public on matters of public interest; the obligation does not apply if the content has undergone human review and a named person holds editorial responsibility. Article 50(5) requires notices to be clear, distinguishable and given at the latest at the time of first exposure.
Whether this blog falls under the rule at all is debatable; the exception for editorially controlled content would probably apply. I label everything anyway, consistently. Not because I have to, but because a blog about AI systems shouldn’t be cagey about how it’s made itself.