The EU’s AI transparency obligations became applicable on 2 August 2026, and the most consequential detail is not the marking requirement itself but the exemption attached to it.
What actually changed
Article 50 of the EU AI Act sets transparency obligations in four situations. The two that matter for anyone producing content are these.
Providers of generative AI systems must ensure that synthetic output is marked in a machine-readable format and detectable as artificially generated. Systems already on the market before 2 August 2026 have until 2 December 2026 to meet the marking requirement.
Deployers publishing AI-generated text intended to inform the public on matters of public interest must disclose that the content is artificially generated.
The obligations apply to providers and deployers of any AI system used in those situations, not only to systems classified as high risk, and they extend to open-source systems.
The exemption that decides most cases
The disclosure obligation on public-interest text does not apply where the content has been subject to human review or editorial control and a natural or legal person holds editorial responsibility for the publication.
There is a second, narrower exemption. The marking obligation does not apply where the AI performs only an assistive function for standard editing, such as grammar correction, and does not substantially alter the input.
Read those together and a principle falls out.
The line between labelled and unlabelled is not how the words were produced. It is whether an identifiable person reviewed them and can be held responsible for them.
That is the first serious attempt any regulator has made at defining what makes AI-assisted work yours, and it is a better definition than most of the ones circulating in marketing.
Why this matters even outside Europe
Three reasons.
Rules of this kind get copied. Whatever your view of the regulation, standards travel, and the editorial responsibility test is a portable idea that other jurisdictions and platforms can adopt without adopting the whole act.
It gives you a defensible internal policy. If you are trying to write a rule for your own team about when AI-assisted content needs a disclosure, “did a named person review this and take responsibility for it” is a cleaner test than trying to measure how much of a draft a model wrote.
It reframes the AI slop argument. Most of the anxiety about AI-generated content is really anxiety about content nobody took responsibility for. The regulation has landed on that distinction rather than on word origin, and that is the more useful place to draw the line.
Provider obligations and deployer obligations are different
It is worth separating the two, because most marketing teams are deployers rather than providers.
A provider builds or places a generative system on the market. The marking obligation sits with them, and it is technical: output has to carry a machine-readable indication that it was artificially generated.
A deployer uses the system. The disclosure obligation that touches content teams sits here, and it is narrower than the marking one. It applies to AI-generated text published to inform the public on matters of public interest, and it is the obligation the editorial responsibility exemption applies to.
So a marketing team publishing product pages is in a very different position from one publishing news commentary, and the practical question is which of your content types would count as informing the public on a matter of public interest.
What this does not do
It does not tell a reader, or a search system, whether a piece of coverage was earned, pitched, contributed, placed, sponsored or paid for. A machine-readable mark tells you a machine wrote it. It says nothing about who decided it should exist or who paid for it to.
Those are different questions and only one of them now has a rule attached.
A practical checklist
Identify who holds editorial responsibility for each content type you publish. If the answer is nobody, that is your first gap regardless of jurisdiction.
Separate assistive editing from generation in your own workflow documentation. Grammar correction and substantive drafting are treated differently and you should be able to show which is which.
Check what your tools emit. If you use a generative system whose provider marks output in a machine-readable format, that marking may travel with the text into places you did not anticipate.
Do not confuse disclosure with provenance. A piece of content can be fully compliant with the marking rules and still tell a reader nothing about how it came to be commissioned.
FAQs
When did the EU AI Act transparency rules take effect?
The Article 50 transparency obligations became applicable on 2 August 2026. Generative systems already placed on the market before that date have until 2 December 2026 to meet the machine-readable marking requirement.
Does all AI-generated content need to be labelled?
No. The marking obligation applies to synthetic output from generative systems, and the separate disclosure obligation covers AI-generated text published to inform the public on matters of public interest.
What is the editorial responsibility exemption?
AI-generated public-interest text does not require disclosure where it has undergone human review or editorial control and an identifiable person or organisation holds editorial responsibility for the publication.
Is grammar checking covered by these rules?
No. Where an AI system performs only an assistive function for standard editing and does not substantially alter the input, the marking obligation does not apply.
Do these rules apply to companies outside the EU?
They apply where systems and their output reach the EU market. Beyond legal scope, the editorial responsibility test is a portable standard that other jurisdictions and platforms may adopt.
Does a machine-readable AI mark tell you who paid for content?
No. Marking indicates that content was artificially generated. It carries no information about commissioning, sponsorship, or how the piece came to be published.
ABOUT AXIS SUITE
Axis Suite by TrendAxis is the independent intelligence layer that explains what AI believes about your brand, why it believes it, and what decision that belief ultimately drives.
Most tools count citations. Axis Suite tells you which of those citations are actually evidence about you, how many separate parties they represent, and what would remain if your largest source disappeared.
Run a scan at axissuite.ai or read the methodology in the Proof Center.