Why I'm building this
I've spent the last two years building AI features for a living. Chat assistants, image generation, streaming LLM stuff.
Last month I started reading the EU AI Act properly, and one part surprised me.
Since 2 August, any AI chatbot used by people in the EU has to tell them they're talking to an AI.
And by 2 December, AI tools that were already on the market have to label what they generate. Images, audio, video and even text. Not a visible watermark, a machine-readable label that software can detect.
It applies to companies outside the EU too, if they have EU users.
When I tried to figure out how to actually do this in code, I found a mess. Different libraries for different file types. Labels that get stripped the moment an image goes through a CDN. No real standard for text at all.
So I'm building Label50. One npm install to disclose, label and keep proof of every AI output.
Before I write much code, I want to hear from people who ship AI products. If you have EU users, how are you handling this? Or is it not on your radar yet?
Genuinely curious. Comments or DMs both work.
https://label50.dev/
Label50Label50 – EU AI Act Article 50 SDK for AI appsDisclose, label and prove every AI output in one npm install. Chatbot disclosure, C2PA labels and evidence for the EU AI Act deadline of 2 Dec 2026.
commit a12ecb5