EU AI Act Article 50 Compliance Kit — chatbot disclosure & AI content labeling, done in a day

Does this apply to you? (60-second check)

Your app has a chatbot or AI agentArticle 50(1): users must be informed they're interacting with AI — unless it's already obvious. You need compliant disclosure at the right moments, not a buried terms-page line.

Your app generates images, video, audio, or textArticle 50(2): outputs must be marked as AI-generated in a machine-readable way. Marking has a transition window for existing systems — but the duty starts now.

You publish AI content or deepfakesArticle 50(4): visible labeling duties for deployers, with narrow carve-outs (artistic works, human-reviewed text) you must apply correctly and document.

You're not in the EUNeither are we. The Act reaches any provider whose system's output is used in the EU — a Hong Kong, US, or Singapore app with European users is in scope.

What's in the kit

Start-here triage — a decision tree that tells you in 10 minutes which of the four Article 50 duties apply to your product (provider vs deployer, with the exemptions mapped).
Disclosure copy library — paste-ready chatbot/agent disclosure strings for first-message, persistent-UI, and voice contexts, in English, short-form for all EU languages, and Chinese — with placement guidance built on the regulation's "obviousness" standard.
Marking verification & implementation guide — per-modality machine-readable marking (C2PA content credentials, SynthID, metadata approaches): what OpenAI, Google, and Anthropic now emit upstream, the survival tests that prove those marks make it through your resize/re-encode/CDN pipeline intact, and exactly what your app must add itself.
Deepfake & AI-text labeling patterns — visible label components and caption templates for 50(4), plus the human-review carve-out documented properly.
Evidence log — the compliance register that shows what you implemented, where, and when. If anyone ever asks, you open one file.
Code-of-Practice & Guidelines mapping — every artifact cross-referenced to the June 2026 Code of Practice on marking and the Commission's guidance, updated as the draft guidance finalizes.
Edge-case FAQ — AI-assisted editing, internal tools, B2B products, obviously-artistic content, upstream-marked media.

PDF + editable filesEN + 中文Updated through the 2 Dec 2026 marking wave

$190 one-off

Rises toward $290 as the 2 December marking deadline approaches. Optional update tracker (+$49/mo): guideline changes, Code-of-Practice updates, and the December 2 countdown. Kit current to August 2026 — final Guidelines, approved Code of Practice, and the Commission's official AI-label icons are all reflected.

Get the kit — instant download

14-day unconditional refund. If it doesn't save you a week, keep your money.

Questions we'd ask too

Is this legal advice?

No. This is an implementation kit — organized information and ready-to-use artifacts built from the regulation, the Code of Practice, and the Commission's guidance, each with source citations. For legal questions about your specific situation, consult counsel; the kit will make that conversation shorter and cheaper.

The big AI Act deadlines got delayed — doesn't that include this?

No. The 2026 Digital Omnibus postponed the high-risk system obligations to 2027–2028. Article 50's transparency duties were not postponed: they apply from 2 August 2026, with a transition window only for the machine-readable marking of systems already on the market.

My app just wraps OpenAI/Gemini/Claude — am I covered by their marking?

No — and 2026 made this trap sharper in both directions. OpenAI ships C2PA + SynthID on images and Google ships SynthID today; Anthropic announced text watermarking + C2PA in August but it isn't live yet. Relying on "my provider handles it" fails one way when marks get stripped in your pipeline — and the other way when they were never there. The final Guidelines settled that the application-layer provider owes the duty. The kit's upstream checklist confirms what your provider actually emits, the survival tests prove what reaches your users, and the evidence log records both.

What are the penalties?

Transparency-obligation breaches carry fines set by the Act's penalty regime (the kit states the exact tier with citations). The bigger near-term risk for small vendors is enterprise customers demanding proof of compliance in procurement — the evidence log answers both.

Why isn't this free like the checklists?

Free checklists tell you the law exists. This hands you the sentences, the metadata implementation steps, and the register — the parts that otherwise cost a week of your time or four figures of counsel time.