AI Disclosure Kit - Ship EU AI Act Article 50 disclosures in 60 seconds
by•
Generate compliant EU AI Act Article 50 disclosures for chatbots, AI-generated content, deepfakes, and biometric systems. Free to try. Deadline 2 August 2026.
Replies
Best
Hunter
📌
Hey Product Hunt 👋
I'm Andrzej, and I built AI Disclosure Kit because I got tired of Googling Article 50 every time a client asked me "wait, do we need to tell users the chatbot is AI?"
The backstory: the EU AI Act enters full enforcement on August 2, 2026. Article 50 specifically requires any AI system interacting with humans (chatbots), generating synthetic content (image/video/audio/text), doing emotion/biometric recognition, or creating deepfakes to tell users it's AI. Fines: up to €7.5M or 1.5% of global turnover. Interesting coincidence: the EU Digital Omnibus trilogue is finalizing around now — if the rumor is true, watermarking obligations slip to Feb 2027 but user-facing disclosure stays at Aug 2. Either way, this tool matters.
Most EU SMEs aren't ready. Most are stuck in a "Google the Regulation → find a €30k/yr enterprise compliance suite → push it to Q3" loop.
What this tool does:
- Pick your AI system type (chatbot, image gen, deepfake, emotion recognition, etc.)
- Pick industry + language
- Get disclosure text grounded in the exact Article 50 clauses with citations (50(1), 50(2), 50(3), 50(4))
- Copy-paste HTML banner + compliance-file note for your audit trail
- 10 EU languages, 3 free generations per day, no signup
What it's NOT: legal advice. Output is a starting template — have counsel review before production. We're upfront in Terms.
Tech (for the PH builder crowd): Cloudflare Workers single-file, D1 + KV, OpenAI gpt-4o-mini (~$0.00042/gen). 92 KiB / 24 KiB gzip. LCP sub-200ms. No framework. Jurisdictional text is a curated prompt, not hallucinated. Built solo with Claude Code as pair-programmer in 3 weeks.
Pricing: Free 3/day · Pro €29/mo unlimited + API + audit log · Team €99/mo CI/CD webhook + 5 seats.
Feedback especially welcome from EU compliance practitioners on the clause grounding — if you see anything imprecise in the Regulation mapping, please call it out.
→ Try it: https://disclosekit.com
Report
Hunter
A question already came up this morning from an early tester worth sharing:
"Does this cover a support chatbot on our Contact page? It just answers FAQs."
Short answer: yes, Article 50(1) applies.
The obligation triggers on any AI system that interacts with a natural person — the use case doesn't matter. Support bot, booking assistant, "just FAQ," intake form with AI autofill — all in scope if a human is on the other end and the system uses AI to generate responses.
Exception: if it's "obvious from the context" to a reasonably informed person that they're talking to AI. In practice that exception is narrower than most teams assume, and the burden of proof is on the deployer.
Kit generates the disclosure text tuned to your specific context (chatbot vs. synthetic content vs. emotion recognition vs. deepfake) in the 10 main EU languages. Free tier gives you 3 generations — enough to test the flow end-to-end before deciding.
Replies
A question already came up this morning from an early tester worth sharing:
"Does this cover a support chatbot on our Contact page? It just answers FAQs."
Short answer: yes, Article 50(1) applies.
The obligation triggers on any AI system that interacts with a natural person — the use case doesn't matter. Support bot, booking assistant, "just FAQ," intake form with AI autofill — all in scope if a human is on the other end and the system uses AI to generate responses.
Exception: if it's "obvious from the context" to a reasonably informed person that they're talking to AI. In practice that exception is narrower than most teams assume, and the burden of proof is on the deployer.
Kit generates the disclosure text tuned to your specific context (chatbot vs. synthetic content vs. emotion recognition vs. deepfake) in the 10 main EU languages. Free tier gives you 3 generations — enough to test the flow end-to-end before deciding.