Does your website have an AI chat, AI-generated images, or AI-written content? The EU is tightening the rules on showing that AI is doing the talking or the creating. Here is what applies to you, and what does not yet.

The EU says 2 August. Norway does not, quite yet.

The European Commission published guidance on 20 July on the transparency obligations in the EU AI Act. The obligations start applying inside the EU from 2 August 2026.

Norway is not an EU member, it is an EEA member. EU law does not apply automatically here. It first has to be incorporated into the EEA Agreement and passed as Norwegian law, the planned "KI-loven". That process is not finished, and according to kiforordning.no, Norway's own entry into force may slip to around 2027, partly because the EEA negotiations are taking longer than expected.

So no, there is no Norwegian legal deadline on 2 August 2026. Not yet, anyway.

Why it can still reach you

The AI Act uses the same trick as GDPR: it does not stop at the border. Article 2 of the Act states that the rules also apply to providers outside the EU where "the output produced by the AI system is used in the Union". If your chatbot answers a visitor sitting in Germany or France, the Act can already apply to that specific conversation, regardless of whether Norway has adopted its own version yet.

Most Norwegian websites have visitors from other European countries, if only through search, ads or a newsletter. So "we will wait until Norway is done" is not a safe strategy in practice, even though it is true that no Norwegian legal deadline exists today.

Two things hit an ordinary business website directly once the rules apply:

  • AI chatbots must disclose that they are AI, unless that is obvious from context.
  • Certain AI-generated or AI-manipulated content (images, audio, video) must be labelled or made machine-readable, especially content that resembles real footage or looks like journalism.

What it means for you

Do you have a chatbot on your website? Check that it introduces itself as AI, not as "Kari from customer service". It is not enough that a disclosure sits somewhere in the terms nobody reads.

Do you use AI-generated images or video in marketing? If someone could reasonably mistake it for a real photo or real footage of a person, it should be labelled.

Do you publish articles or news where AI did much of the work? Then the point is not only the law. It is trust. A person should stand behind the content, be able to answer for it, and have read it before it goes out.

None of this means you have to stop using AI. It means being honest that you use it, and having a simple routine for who is accountable. For most small businesses this is a couple of hours of work, not a project.

What happens if you leave it

For most small Norwegian businesses, the immediate risk is not a fine in tomorrow's post, from Norway or the EU. It is enforcement that builds up gradually as the KI-loven and EEA incorporation fall into place, and it is customers who lose trust once they discover that "customer service" was a bot the whole time.

Documentation is cheap insurance. If you have written down how you use AI and how you disclose it, you have something concrete to point to the day someone asks, whether that is a customer, a journalist or a regulator.

How we do it at Ibenta

Our own AI chat on ibenta.no discloses itself the moment the chat window opens: it is labelled "AI assistant", not a name pretending to be a person. That is not a last-minute fix for the EU deadline. That is how we built it.

We have also written down how we work editorially with AI in content on ibenta.no: when AI is used for research and drafts, what a person always reviews before publishing, and how corrections are handled.

Check this today

  • Open the chatbot on your website. Does it disclose that it is AI, clearly, right away?
  • Look through your AI-generated images and video. Could anyone mistake them for real footage?
  • Check whether anyone actually owns your content editorially, or whether it just "gets published".
  • Write your practice down somewhere, even briefly. Then you have something to point to if someone asks.

Go through these four checks yourself, it takes ten minutes. Unsure whether your chatbot or content workflow actually meets the requirement? Get in touch, we will look at it together, no obligation.