Thirteen days from now, on 2 August 2026, Article 50 of the EU AI Act becomes enforceable. It’s easy to have missed this, because on 29 June the Council of the EU gave final approval to a “Digital Omnibus” package that pushed the Act’s higher-profile high-risk AI system obligations back 16 months, to 2 December 2027 — and a lot of the coverage since has read as “the AI Act got delayed.” Article 50’s transparency duties weren’t part of that delay. They land on schedule.
What Article 50 actually requires is narrower than the Act’s headline risk rules, but it’s not trivial. Any AI system that interacts directly with people — a chatbot, an AI assistant — has to make clear to users they’re dealing with AI, unless that’s already obvious. Generative AI outputs — text, images, audio, video — have to be marked in a machine-readable format that’s detectable as AI-generated, though systems already on the market before 2 August get a grace period to 2 December 2026 to meet that specific marking requirement. Anyone publishing AI-generated deepfakes or manipulated content has to disclose it clearly. Penalties for non-compliance sit in the same tier as the Act’s other serious breaches: up to €15 million or 3% of global annual turnover, whichever is higher.
The part most UK founders get wrong is scope. The EU AI Act reaches any provider or deployer whose AI system’s output is used by people in the EU — it doesn’t matter whether you have an EU entity, an EU office, or even an EU customer contract. If your product ships an AI chatbot, AI-generated copy, or an AI image or content tool, and any of your users are in the EU, you’re in scope for Article 50 regardless of where your company is registered.
So what
This is a genuine “it depends” situation, not a blanket panic. If you’re building an internal tool or something with a purely UK/domestic user base, Article 50 isn’t your immediate problem. But if you’re commissioning an AI-powered product — a support chatbot, a content generator, anything customer-facing with generative AI in the loop — and there’s any realistic chance of EU users, the disclosure UX and machine-readable content marking need to be designed in now, not retrofitted after a compliance letter arrives. That’s a genuinely different build to one where AI transparency is an afterthought: it touches UI copy, output metadata, and how the product discloses AI involvement by default. If you’re scoping an AI product and aren’t sure where you land on this, our AI products work covers exactly this kind of build, or get in touch and we’ll help you work out what actually applies to you.