The editorial exception in Article 50(4): when AI text needs no label

When AI-generated text needs no label: the conditions in Article 50(4), the Commission's guidelines and Commitment 4 of the EU Code.

Zespół AIMARKER· 10 October 2026· 4 min read

A document with a green approval check and the AI icon

Article 50(4) of the AI Act requires AI-generated or manipulated text published to inform the public on matters of public interest to be disclosed [1]. The same provision contains an exception for text that a human has reviewed. Here is when it applies and how to document it.

The wording

The obligation does not apply "where the AI-generated content has undergone a process of human review or editorial control and where a natural or legal person holds editorial responsibility for the publication of the content" [1, Art. 50(4), second subparagraph].

Both conditions must be met:

human review or editorial control,
a natural or legal person holding editorial responsibility for the publication.

The exception covers text only. For deep fakes (images, audio, video), the provision only limits the obligation for evidently artistic, creative, satirical or fictional works; it does not exempt them [1, Art. 50(4), first subparagraph].

What counts as editorial control?

The European Commission's guidelines on Article 50 set out the requirements [2]:

human review means a deliberate examination of the substance by one or more people with relevant knowledge and professional judgement in the subject, and fact-checking the accuracy of the content is a minimum requirement [2, para. 134],
not sufficient: spelling and grammar checks, the mere existence of an editorial policy, automated review processes, or cursory editorial approval [2, para. 135],
any substantive AI intervention after the human review or editorial control process makes the exception void [2, para. 136],
editorial responsibility means ultimate legal responsibility for the publication, and the identity and contact details of the person, entity or function holding it should be publicly available in an easily findable location: online e.g. in the terms and conditions or other legal information, in print in the colophon [2, para. 138].

What does the Code of Practice say?

The Code of Practice on Transparency of AI-Generated Content of 10 June 2026 has a dedicated commitment on this exception [3, Commitment 4, pp. 35–36]. Signatories that are not media service providers keep a policy covering, among other things:

the identification of the natural or legal person with editorial responsibility (name, role and contact details),
an overview of the organisational measures and human resources that ensure review or editorial control before publication; the Code notes this does not entail documenting each individual review.

Where not already publicly available, signatories commit to publishing the contact details of those with editorial responsibility [3, p. 36]. Media service providers within the meaning of the European Media Freedom Act that are subject to editorial standards may rely on their existing procedures [3, p. 35].

How to document it: our recommendation

Neither the Regulation nor the Code requires documenting each individual review. Such records do make it easier to show the conditions were met, e.g. during an inspection or with a client. As good practice, we suggest recording for each text:

that AI was used and which tool,
who carried out the review and when,
what the review covered, including checking facts and sources,
who holds editorial responsibility and where their contact details are published,
whether the text was changed with AI after the review.

Specialist content

The guidelines require the review to be done by people with relevant knowledge and professional judgement in the subject [2, para. 134]. For texts on health, finance or law, this in practice means e.g. a doctor or a lawyer. Errors in such texts also have more serious consequences for readers.

Exception or label?

The exception does not prohibit labelling. You can label texts voluntarily even when the conditions are met, e.g. "Prepared with AI, reviewed by our editors". It is transparent for readers and reduces the risk of disputes over whether the review was sufficient.

How AIMARKER helps

Editorial review in the register: for each text you record who reviewed it and when.
Approval by link: a client or specialist approves content without an account; we record their name, date, IP address and a copy of the content.
"Does it need a label?" wizard: questions based on Article 50, with the reasoning saved in the report.
An audit log whose entries cannot be changed or deleted, and a PDF report.

See how it works.