Are you treating the Aug 2 AI Act deadline as real?
Genuine question for anyone shipping an AI feature to EU users.
On Aug 2 the AI Act's transparency rules (Art. 50) start applying — and for most indie apps it's tiny: say it's an AI, label AI-generated media. Not the "you need a lawyer" monster people fear.
But I keep going back and forth on how seriously builders should take it right now, given enforcement will realistically take time to reach small players.
So I'm curious where this community lands:
— Are you doing anything before Aug 2, or waiting to see how enforcement actually plays out?
— What's the fuzziest part for you — the transparency labels, or figuring out if you've accidentally drifted into a "high-risk" use case (hiring, credit, scoring)?
Happy to share what I know in the comments if it helps.
Replies
I'm leaning toward implementing the basics now. Even if enforcement takes time, being transparent with users feels like the right direction anyway.
@debra_salt That's the healthiest framing I've seen in this whole debate, honestly — the regulator sets the deadline, but the user is the actual reason. And the trust math works in your favor: people react far better to "yes, this is AI" said upfront than to discovering it later. The disclosure isn't a legal tax, it's just… not surprising your users.
Since you're leaning toward doing the basics now, here's what "the basics" actually means for most apps — it's genuinely an afternoon:
One honest line in the UI where users interact with the AI ("responses are AI-generated")
If you generate realistic images/audio/video — label it as synthetic
Name your model provider in the privacy policy (this one's less AI Act, more GDPR — but it's the same afternoon)
That's the whole list unless you're in hiring/credit/scoring territory. And the nice side effect: you need #1 for FTC reasons in the US anyway, so it's one fix covering two jurisdictions.
Curious what you're building — that usually decides which of the three matters most for you.
@debra_salt @ilyaofliian I agree with treating transparency as worth doing on its own terms. Our situation is slightly different - our product doesn't have AI features, but we use AI for coding and to generate some images (not realistic, more like illustrations) and marketing visuals. So for us the question isn't whether to disclose in-app AI, but whether an AI-generated image on a marketing page needs the same labeling.
@debra_salt @alieksia
Good news: your case is mostly outside Art. 50.
The labeling duty targets realistically generated content — the deep-fake zone. Things that look like a real photo of real people, places or events. Illustrations and stylized marketing visuals don't trigger it.
Nobody has to caption a cartoon.
AI for coding is also irrelevant here. That's a dev tool, not an AI feature in your product. The Act cares about what you ship, not how you wrote it.
Two places where it flips back on:
If you generate photorealistic images that could pass for actual photos — fake team photos, fake customer shots, "product in use" scenes that never happened. That's disclosure territory. And honestly, US FTC deception rules bite there too, AI or not.
If you later add an AI feature users actually interact with — then the one-line disclosure applies.
So: illustrations — post freely. Photorealism pretending to be photography — label it or don't use it. That basically it :)