What must you tell users when AI is involved? Draft it before launch, not after.

There's a moment every AI feature reaches: a user realizes a model made a decision, and asks why. If your answer isn't already written down, you're not just behind — you're in the exact gap the EU AI Act's transparency duties are aimed at.

I built DiscloseKit because "we should probably tell users" is a sentence that never survives contact with a launch deadline.

30-second takeaway

• Transparency isn't a banner you bolt on later; it's a statement you should draft while the feature is still a plan.
• The hard part isn't writing — it's knowing which interactions actually trigger a duty.
• A draft you can review beats a policy nobody wrote.

The disclosure nobody schedules

Most teams know they should tell users when AI is involved. Almost none put it on the build checklist. So it lands in a frantic rewrite after a complaint, a journalist, or a regulator asks the obvious question. By then the copy is defensive, not clear.

Drafting the disclosure early — from the real interaction, not a template — means the words are honest because they were written before anyone had a reason to spin them.

What the tool actually does

You describe where AI sits in your product. It drafts the disclosure language and points at the interaction that triggers a duty, so legal gets a starting point instead of a blank page. The output is a draft you own and can argue with — not a compliance certificate.

DiscloseKit runs in the browser, no account, no key. Describe the touchpoint and get a disclosure draft.

FAQ

What is DiscloseKit?

A tool that drafts the transparency disclosure you owe users when AI is involved in a product interaction. No account, no key.

Does a draft make me compliant?

It gives you the words to start from. Whether they're sufficient is still a human call — this is decision-support, not a sign-off.

Is this legal advice?

No. It drafts from your description; it doesn't interpret the law for your situation.

Author's Process

This product was built with AI coding tools. The interesting part was resisting the urge to over-claim — the value is a clear draft, not a false promise of safety.