Why this blog now carries an AI badge
Starting today, a small badge will mark every post whose main text was significantly co-written by AI. The AI Act doesn't require me to do this - I'm doing it anyway.
Machine-translated from German, reviewed by the author. German original
In my last post about sovereign LLM architectures, I noticed how much of the text actually didn’t come from me. The structure, the core theses, the technical decisions - all mine. But entire paragraphs in the final wording were written by an AI; I only approved and checked them. I no longer wanted to silently leave it that way.
What the badge shows
Every post in which a model contributed to the wording now carries a small badge next to the date and tags: “IDEA.” Spelled out: I Direct, Engine Assembles - the idea, concept, and technical context come from me; the model only assembled the text. A tooltip explains this when you hover or focus:
I Direct, Engine Assembles — the idea, concept and thinking are mine; the prose was drafted with AI assistance.
Technically, it’s a single flag evaluated by my blog engine - per language version, not per post. A post may be handwritten in the original and carry the badge only in translation, or vice versa.
What the AI Act requires of me - namely, nothing
The part that amused me when I looked it up: Article 50(4) of the AI Act mandates labeling AI-generated text on matters of public interest - but only if the content hasn’t undergone human review and no natural or legal person bears editorial responsibility. I read every sentence before publication, correct, shorten, contradict myself, and stand behind every claim in the text with my name. That puts me outside the scope of the regulation before it even begins.
I’m still adding the badge anyway. Not because a law demands it, but because readers here have a right to know whose handwriting they’re reading - even if legally, no one cares except me.
What doesn’t change as a result
Every post starts with an outline I write myself: bullet points, core theses, code examples that actually come from me and aren’t fabricated. What an AI takes over is exclusively the final mile - smoothing sentences, removing repetitions, producing the target text in translations. Before publication, I read every post in full, change what sounds wrong or doesn’t match my voice, and stand behind every line professionally. The badge changes nothing; it only makes visible where in the process the AI contributed.
For a blog that writes about AI architectures more often than the average competitor, it would have been inappropriate to remain opaque about where one’s own text comes from. Anyone in their own company who will eventually have to disclose which content was created with AI assistance is welcome to adopt this principle for free.