Your AI features become illegal in the EU
on August 2, 2026 — unless you can prove they're compliant.
Every chatbot, every AI-written line, every generated image or voice now has to be labelled, watermarked, and documented under EU law. Buyers already ask for the paperwork in procurement — and deals die when you can't produce it. ClearMark gives you both: the technical marking and the compliance pack, in days, not months.
Three things happened at once. Together they make this unavoidable.
It's law, with a hard date
Article 50 of the EU AI Act forces anyone whose product uses generative or interactive AI to label and machine-mark its outputs. It's enforceable August 2, 2026 — not a guideline, a regulation.
Buyers now demand the paperwork
AI-compliance documentation is now its own section in enterprise RFPs and security questionnaires. The blunt reality from the field: "if you can't immediately produce the docs, the deal is dead."
There's no standard tool yet
The EU hasn't even finalized the technical watermarking standard. So every company is stitching together immature techniques alone — and most have nothing in place at all. That's the gap.
Sources: EU AI Act Article 50 (artificialintelligenceact.eu) · European Commission regulatory framework · Gartner (AI regulation to reach 75% of the world's economies by 2030).
Find out exactly what Article 50 requires of you.
Answer a few questions. We'll tell you which obligations apply, your risk level, and precisely what's missing — in plain English, not legalese.
Everything Article 50 asks for. Generated for you, ready to ship.
Two layers: the technical marking that makes your AI outputs legally detectable, and the documentation that makes procurement say yes.
⚙️ The technical layer
- Output marking SDK — embeds machine-readable C2PA provenance into your generated images, audio, video & text so they're detectable as AI, exactly as 50(2) requires.
- "You're talking to an AI" disclosure kit — drop-in UI snippets that satisfy 50(1) for chatbots and assistants.
- Deepfake / synthetic-media labels — compliant disclosure components for 50(4) content.
- Verification endpoint — let anyone confirm a piece of content carries your valid mark.
📋 The documentation layer
- Article 50 transparency statement — auto-generated from your answers, in the language buyers expect.
- Procurement-ready compliance pack — the PDF you drop straight into an RFP or security questionnaire.
- Obligation map — which clauses (50.1–50.4) apply to which of your features, with evidence.
- Audit trail — timestamped record proving when and how you complied.
From "we have no idea" to compliant — in three steps.
Check & map
Run the free exposure check. ClearMark maps every Article 50 obligation to your actual AI features and shows your gaps.
Mark & document
Drop in the marking SDK and disclosure snippets. ClearMark generates your transparency statement and audit trail automatically.
Hand it over
Export the procurement pack. When a buyer's questionnaire asks "is your AI compliant?", you attach one PDF and move on.
If your product touches AI and any EU user, this is you.
The honest, complete FAQ.
Does this actually apply to me if I'm not in Europe?
Yes — if any of your users or customers are in the EU. The AI Act applies based on where your users are, not where your company is. A US-based SaaS with EU customers is fully in scope. That's most growing software companies.
What exactly does Article 50 require?
Four transparency duties: 50(1) tell people when they're interacting with an AI; 50(2) mark AI-generated outputs (text, image, audio, video) in a machine-readable, detectable way; 50(3) disclose emotion-recognition / biometric categorization; 50(4) label deepfakes and AI-generated public-interest text. ClearMark covers all four — the free check tells you which apply to you.
What's the penalty if I ignore it?
Breaching the transparency obligations can cost up to €15 million or 3% of global annual turnover, whichever is higher. But for most companies the faster, more certain pain is commercial: enterprise buyers walking away because you can't produce compliance documentation.
Isn't the standard still being finalized? Why build now?
That's exactly why now. The legal obligation lands August 2, 2026 even though the detailed technical standard is still in draft. ClearMark builds on C2PA — the provenance standard the official rules are converging toward — so you're compliant today and aligned with where the spec is heading. Waiting for "final clarity" means missing the deadline.
Is there a grace period?
Partly. AI systems already on the market before Aug 2, 2026 get extra runway (into Dec 2026) under the provisional AI Omnibus agreement — but anything newly launched must comply from Aug 2. If you're shipping new AI features this year, the clock is now.
Can't I just write the disclosure myself?
You can write "this is AI" on a page — that's the easy 10%. The hard part is the machine-readable marking of every output (50.2) and the documented evidence a buyer's procurement team will actually accept. That's the orchestration ClearMark automates so you don't burn an engineer-month on it.
Will this slow down or change my product?
No. The marking SDK adds invisible provenance metadata to outputs and lightweight disclosure UI — no change to your model, your latency, or your UX beyond a small "AI" indicator where the law requires one.
Where is my data hosted?
EU-resident by default. Given the parallel push on EU data sovereignty, we keep processing inside the EU so the compliance tool itself isn't a new compliance problem.
I'm a tiny team / solo founder. Is this overkill?
The opposite — it's built for you. Big companies have legal departments; you have a deadline and a deal at risk. ClearMark is the fastest path from exposed to covered without hiring anyone.
Be compliant before the deadline — not scrambling after it.
Join the early-access list. You'll get the exposure check results, the marking SDK, and the procurement pack first — plus a deadline checklist so nothing slips.
No spam. One launch email + your deadline checklist. Unsubscribe anytime.