Does it apply to UK companies serving EU customers?

Yes. The Act applies to actors outside the EU too, if their output is used within the EU — and that includes UK companies, which have been outside the EU since Brexit.

The trigger is where your AI system's output lands, not where your company is headquartered or incorporated. A UK-based company with EU customers, EU website visitors, or an EU-facing chatbot or content workflow is in scope on the same basis a US or any other non-EU company would be — UK status specifically confers no exemption.

This applies to the same two duties that matter most for SMEs: if your branded chatbot is reachable by people in the EU, the Article 50(1) disclosure duty likely applies to that interaction. If you publish AI-generated content that EU audiences see, the Article 50(4) labelling duties can apply on the same logic.

If your EU exposure is limited — a handful of overseas customers rather than a genuine EU market — that's still worth checking rather than assuming away, since the rule doesn't have a minimum-scale threshold built in. The free check starts with exactly this question ("Do you have customers, users, or website visitors in the EU?") and routes from there.

It's also worth separating this from the UK's own domestic AI regulation, which is a distinct framework covering UK-facing activity. That the UK has its own approach doesn't create a reciprocal exemption from the EU AI Act — a UK company's obligations under the EU Act are determined purely by whether its output reaches EU-based people, independent of what rules apply to it domestically.

A concrete example: a UK-based SaaS company with a knowledge-base chatbot that free-trial signups from Germany or France can use is, on the territorial test alone, in the same position as if that chatbot were built by a company in Berlin. Nothing about being a UK company, post-Brexit, changes that specific analysis — only where your users are does.