AI Now Has to Tell You It's AI. Just Not in Norway.

On 2 August, a right you have never had took effect across the EU: the right to be told that the thing you are talking to is not a person.

Not in Norway, though. We'll get to that.

What the rule actually says

Article 50 of the AI Act is the transparency chapter, and it is short enough to summarise honestly.

If you interact with an AI system, the provider has to make sure you know — unless it is obvious to a reasonably well-informed person. The chat window on a customer service page has to own up. So does a voice on the phone that has been built to sound like it isn't one.

If a system generates synthetic audio, images, video or text, the output has to be marked in a machine-readable format, so it can be detected as artificial further down the line.

And if someone publishes a deepfake, they have to disclose it. Same goes for AI-generated text published on matters of public interest — unless a human reviewed it and someone carries editorial responsibility for it. That last clause is the one every newsroom in Europe read twice.

Getting it wrong carries fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher.

The part that quietly slipped

Six days before the deadline, the EU changed its own schedule. Regulation (EU) 2026/1744 — the Digital Omnibus on AI — was published on 24 July and entered into force on 27 July. It moved the high-risk obligations, the expensive ones industry has been complaining about since 2024, from this August to 2 December 2027. AI built into regulated products got until August 2028.

The transparency date did not move.

That is the interesting bit. The loud, industrial half of the AI Act was deferred by sixteen months. The half that affects you personally, on an ordinary Tuesday evening, arrived on time.

With one hedge: generative systems already on the market before 2 August have until 2 December 2026 to implement the marking requirement. So the labelling is real, but for the tools you already use, it is a December thing.

Two new prohibitions land on that same December date. AI systems that generate intimate images of people without their consent, and AI-generated material depicting child sexual abuse. Banned outright — no risk assessment, no exceptions. Worth noticing that this took until 2026.

Norway, two days later

On 4 August — two days after the EU rules took effect — digitalisation minister Karianne Tung announced that the Norwegian AI law will go out on a fresh consultation round this autumn, with the ambition of putting it to Stortinget in spring 2027.

The regulation still hasn't been incorporated into the EEA agreement. Until it is, and until Stortinget passes the law, none of the above is Norwegian law. There is no Norwegian authority to complain to. Nkom is lined up as the coordinating supervisor, but that sits in a bill, not in force.

So: in Denmark, Sweden, Finland and Germany, you have had the right to know you're talking to a machine since 2 August. In Norway, you don't. Ask again in 2027.

You'll get it anyway. That isn't the same thing.

Here is the honest part. You will almost certainly see the labels.

Nobody is going to maintain a separate, unlabelled version of their chatbot for five and a half million Norwegians. Complying everywhere is cheaper than running two versions. Around 190 companies have already signed the Commission's voluntary code of practice on transparency, published on 10 June, which is how most of them intend to demonstrate they're following the rules.

But getting something as a side effect of somebody else's market is not the same as having a right to it. A right survives a company deciding, next quarter, that the Norwegian version doesn't really need the label.

The asymmetry only runs one way, by the way. A Norwegian company selling an AI system into the EU is covered by EU law today — law it had no vote in, and cannot appeal to a Norwegian regulator.

So when a service tells you that you're speaking to an assistant, understand what you're looking at. In most of Europe, that is now the law. In Norway, it is a courtesy.

Courtesies get withdrawn.