On August 2, 2026, Article 50 of the EU AI Act took effect. It requires two things: telling people when they're talking to a chatbot or a generative AI system, and labeling AI-generated or AI-altered text, images, audio, and video as such. Six days earlier, on July 27, an amendment called the "Digital Omnibus on AI" also entered into force — but it left Article 50 itself untouched, only granting systems already on the market a grace period, until December 2, on the machine-readable watermark format.
Why this is a story right now
Another study released around the same time shows exactly the problem this rule targets. Pangram, a text-detection company, scanned over a million posts across LinkedIn, Medium, Reddit, Substack, and X, and found that 41% of LinkedIn's longform posts were fully AI-written. Unlabeled AI text passing as human writing already makes up a large share of what people read in their feeds — and Article 50 is aimed at exactly that: not AI use itself, but hiding it.
It applies even if you're not in the EU
Article 50 is triggered by where your readers are, not where you or your business are based. Whether you run an English-language blog or newsletter from Korea, if any of your readers are in the EU, you fall within its scope in principle. Fines top out at €15 million or 3% of worldwide annual turnover, whichever is larger.
The reading: it's not "don't use AI," it's "don't hide it"
What Article 50 bans isn't using AI — it's concealing that you did. There are four exemptions: content that's evidently artistic, creative, or satirical; content that has gone through substantive human editorial review with a named person holding editorial responsibility; systems used by law enforcement; and narrower exceptions specific to each obligation. The "editorial review" exemption requires naming a reviewer and documenting the review process — a bar that's hard for a one-person blog to clear in practice. For most solo creators, simply labeling the content is the easier path than trying to qualify for an exemption.
What's certain and what isn't — the effective date, what must be labeled, the four exemptions, and the fine ceiling are all confirmed and consistent across the European Commission's own pages and multiple law-firm briefings. What isn't known yet is how this law will actually be enforced against individual creators based outside the EU — it just took effect, so there's no enforcement track record. Falling within scope on paper and actually being pursued are two different things. (Checked as of August 4, 2026.)
If your site gets 100 visitors a week
Fines aren't the reason to act first. If AI helped draft a post, add one line somewhere on the page saying so. It costs almost nothing, and telling readers upfront costs less trust than having them figure it out later and feel misled. This site's own issue posts are researched and drafted with AI help, so the same rule applies here too.