
Last Updated: August 5, 2026
Article 50 of the EU AI Act became legally binding on August 2, 2026. It requires businesses to disclose when content is AI-generated or manipulated, so a person encountering it can tell. If you've seen a headline about this and weren't sure it applied to you, you're not alone. Most of the coverage blurred three separate things together: a labeling law, a separate staff-training requirement, and a voluntary code of practice.
TL;DR
A quick note before you keep reading: this post is general information, not legal advice. I'm not a lawyer, I'm a marketer working through the same changes you are. Check the sources linked throughout, and if a real decision for your business rides on this, talk to an attorney. This area is still developing, and details here may shift before the law and its guidance settle.
Article 50 of Regulation (EU) 2024/1689 requires providers and deployers to disclose AI-generated or manipulated content. That matters because it changes what counts as honest marketing: a synthetic spokesperson or an invented scene now needs a label, the same way a paid endorsement needs one. The law does not ban synthetic images, voices, or video, but it requires disclosure.
One exemption is worth knowing. Marking isn't required for standard editing that doesn't substantially change what a person actually provided, so a color correction or crop stays outside this law entirely.
A second exemption applies specifically to text. AI-generated or AI-assisted text published on a matter of public interest (a blog post or press release, for example) doesn't need a label if a real person reviewed the substance of it and holds editorial responsibility for what got published. A spellcheck pass doesn't clear that bar. Actually reading it and standing behind it does.
Two words decide this:
Article 50 binds EU providers and deployers, a reading the Commission's own FAQ on Article 50 confirms directly. The actual statutory trigger, under Article 2 of the Act, is broader than "does your business have an EU office": it reaches you when the output your AI system produces is used in the EU. In practice, that usually looks like being incorporated here, selling into that market, or targeting an EU audience with the exact asset in question. Outside that, the Act itself doesn't make disclosure mandatory for you, whatever the headlines might imply. It's still worth adopting the habit voluntarily. More on why below.
Here's a simple way to check: could the content pass as something real? A photorealistic generated person clears that bar. So does a realistic scene of a storm hitting a real city that never happened. A cartoon, or a dragon flying over a landmark, doesn't, because neither could pass as real in the first place.
If Article 50 reaches your business, the fines are real: up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher. Small businesses and start-ups get a break: the cap flips to whichever figure is lower. Most solo marketers and small shops reading this won't sit in that bracket, but it's worth knowing the number exists before you assume it doesn't apply anywhere.
Article 4 requires businesses using AI tools to make sure their team understands basic risks, like the fact that AI can generate false information with confidence. There's no certification and no dedicated fine attached to it. Separately, the Code of Practice on Marking and Labelling AI-Generated Content, finalized in June 2026, is a voluntary set of best practices, not a legal requirement on its own.
Building a disclosure habit is about having one clear, repeatable way to track which content in your workflow is AI-made, so nothing slips through untagged. Hand-editing in Photoshop stays outside Article 50 entirely, since the law only covers AI systems. An FTC deception claim doesn't ask what tool you used, though, and platforms are already enforcing their own versions of this. Meta labels AI-generated posts across Instagram and Facebook. YouTube requires creators to disclose realistic AI-altered or synthetic content. TikTok runs a similar policy. Whichever rule actually reaches you, the practical answer looks the same.
I built a free tool for exactly this, called ai-transparency.site. Answer a few questions about a piece of content and it walks you through whether it likely needs a disclosure and what that disclosure could say. Treat it, and this post, as a starting point, not a final answer.
In Short
If you're making images, video, or copy with AI and haven't built a disclosure habit yet, this is a good week to start.

Lionel Lowery
I'm Lionel, a marketing and brand strategist who helps brands find clarity and build systems that hold up.
unsure where to start?
Free brand and web audit. No email signup.
Five minutes, a real report, and nothing to unsubscribe from later.