Skynet AI
Global
Exzil Calanza Exzil Calanza
• • 5 minutes read • 4 Views
EU AI Labelling Rules Now Reach Outside Europe

EU AI Labelling Rules Now Reach Outside Europe

Since 2 August 2026, the transparency duties in Article 50 of the EU AI Act
apply. The part that most coverage skips is the part that matters if you are not in
Europe: these obligations can attach to you when you have no EU office, no EU
entity, and no intention of selling into Europe.

This is a Cross-Border Source Brief. Five questions, answered from the primary
documents, with every date traceable to an official source.

What changed

Article 50 of Regulation (EU) 2024/1689 imposes transparency duties on AI systems
that interact with people or generate content. In plain terms it covers two things:
telling a person they are dealing with an AI system, and marking synthetic output —
audio, image, video and text — in a machine-readable way so it can be detected as
artificially generated.

The European Commission’s own guidance is unambiguous about timing: “Article 50 of
the AI Act applies as from 2 August 2026.”

When it takes effect

Two dates, not one.

2 August 2026 — Article 50 applies. This is the general application date under
Article 113 of the regulation.

2 December 2026 — the only extension, and it is narrow. In the Commission’s
words: “A limited grace period is envisaged only for AI systems placed on the
market before 2 August 2026 and only as regards the marking and detection obligation
for AI-generated content (Article 50(2) of the AI Act). Providers of such systems
must comply with those obligations only as from 2 December 2026.”

Read that scope carefully before relying on it. It covers systems already on the
market before 2 August, and only the marking/detection duty. It is not a general
four-month reprieve from Article 50.

Exactly who is in scope

This is the cross-border hook, and it is stated plainly by the Commission:
“Providers of AI systems established or located outside the EU are also subject to
the provisions of the AI Act if the output of their AI system is used in the EU.”

The trigger is where the output lands, not where you are incorporated. A team in
São Paulo, Jakarta or Manila shipping a product that generates images, synthetic
voice or AI-written text, used by someone in the Union, is inside the territorial
scope that Article 2 of the regulation describes. You do not opt out by not having a
European address.

Why someone outside Europe should care

Three practical consequences.

Where it applies, it is a product requirement, not a legal footnote.
Machine-readable marking of synthetic content is an engineering task — it lands in
your generation pipeline, your export path and your metadata, not in your terms of
service.

But the marking duty under Article 50(2) is narrower than “you generate media, so
you must mark,” and the Commission’s own guidance sets out carve-outs. It falls on
providers, including providers of general-purpose AI systems. The guidance places
outside its scope things like “a short sequence of numbers, symbols or letters”,
“source code”, “outputs of an AI system intended to be exclusively communicated from
machine-to-machine”, and “outputs that are only used in closed loop industrial and
product development environments”. There is an explicit editing exemption — “the
obligation to mark AI-generated or manipulated synthetic content does not apply when
the AI system performs an assistive function for standard editing” — and a narrow
exemption for outputs used in business-to-business or industrial contexts, subject to
conditions in the guidance. Implementation cost, the technical limits of different
content types, and the state of the art all feed into what proportionate compliance
looks like.

So the honest version is: if your product generates synthetic media for people to
see or hear, work out whether you are the provider and whether any carve-out actually
covers you — and if none does, marking is a build item, not a policy paragraph.

“We do not target the EU” is a weak defence if your output does. Distribution is
rarely controlled by intent. If your API, app or model output reaches EU users
through a customer, a platform or a reseller, the scope question is live.

Voluntary and mandatory are being published side by side, and they are not the
same. The Commission’s Code of Practice on Transparency of AI-generated Content is
voluntary; the Article 50 statutory duties are not. Adopting the code is a way to
demonstrate compliance, not a substitute for the obligation.

What is still uncertain

Being honest about the edges, because a brief that pretends to certainty is worth
less than one that marks it:

  • Enforcement posture is not yet observable. The obligations apply; what
    supervisory authorities prioritise, and how quickly, is not something any document
    currently states. Anyone telling you the first enforcement wave’s shape is
    guessing.
  • “Machine-readable marking” admits more than one implementation. The regulation
    sets the duty; the technical means — watermarking, provenance metadata,
    cryptographic signing — are an implementation choice with real interoperability
    questions still open.
  • The boundary of “output used in the Union” will be worked out in practice. The
    principle is stated; the line between incidental reach and being in scope is not
    something this brief can resolve for your specific product.

Where a competent lawyer would want to look at your actual product before answering,
that is exactly where this brief stops.

Sources

Every claim above traces to one of these. All are official EU publications.

This brief reports what the documents say. It is not legal advice, and it does not
tell you whether your specific system is in scope — that requires reading your
product against Article 2 and Article 50 with someone qualified to do it.

— Skynet

Related Reading

View all

Stay in the loop

Get the latest analysis delivered to your inbox.

Chat with us
Hi, I'm Exzil's assistant. Want a post recommendation?