From 2 August: Identifying AI Content Is No Longer a Minor Detail

Text type: EDT (Editorial/Classification) | Factual status: 27 July 2026

Today, on 27 July, the European AI Omnibus entered into force. It is intended to streamline procedures in the AI Act, adapt deadlines for individual obligations, and make regulatory sandboxes more accessible. This is an important, yet easily misunderstood moment: simplification does not mean that the early transparency rules vanish.

On the contrary. Already on 2 August 2026, central transparency obligations under Article 50 of the AI Act will become applicable. In a public sphere where texts, images, audio, and video can increasingly be synthetically generated or altered, this is no mere formality. It addresses a very practical question: How do people recognize whether they are interacting with an AI system, viewing synthetic content, or reading an editorially verified publication?

Three Levels to Distinguish

First, there is the law. Article 50 of Regulation (EU) 2024/1689 obliges providers of certain interactive systems to inform natural persons that they are interacting with an AI system—unless this is obvious from the circumstances. For generative systems, the article further mandates machine-readable marking of artificially generated or manipulated content, insofar as technically feasible.

Second, there are obligations for deployers of AI systems. In the case of deepfakes, it must be disclosed that image, audio, or video content has been artificially generated or manipulated. Particularly relevant for public discourse is an additional rule: AI-generated or AI-manipulated texts addressing matters of public interest must be disclosed. Exempted are only texts that have undergone a process of human review or editorial control and where a natural or legal person holds editorial responsibility for the publication. The European Commission published guidelines on this matter on 20 July.

Third, there are implementation tools. The European Code of Practice on Transparency of AI-Generated Content outlines technical watermarks, labels, and procedures for providers and deployers. It is voluntary. The statutory obligations under Article 50 are not. Not signing the Code does not automatically mean non-compliance; compliance, however, must then be demonstrated through other means. Furthermore, a signature does not replace careful case-by-case assessment.

This distinction warrants close attention. A voluntary code can foster common practice and technical interoperability. However, it is neither a carte blanche nor a substitute for intelligible disclosure to real human beings.

What Changes in Practice for Newsrooms and Organizations

The new rules do not serve as a mandate to slap a warning label on every single use of AI. They target specific risks of deception and manipulation. For this exact reason, a precise rather than blanket approach is beneficial.

For newsrooms, this means: An editorially accountable publication does not become unreliable simply because AI aided in research, structuring, or drafting. Crucial for the exemption is that a human review or control process takes place and that a natural or legal person assumes editorial responsibility—including responsibility for sources and corrections. If any of these prerequisites is missing for AI-generated texts on public interest topics, the transparency obligation applies with full force.

For enterprises, associations, educational institutions, and public administrations, a brief inventory is recommended before the deadline:

Map interactive systems: Where do chatbots or virtual assistants handle inquiries from employees, customers, or citizens? Is it clearly intelligible that AI is involved?
Classify content types: Which images, voices, videos, or texts are generated or altered? For which of these might an audience falsely presume human authenticity?
Designate accountability: Who conducts substantive and editorial review of AI-assisted publications? How can errors be reported and corrected?
Evaluate technical verification: Can provenance data or machine-readable watermarks be meaningfully deployed? Visible labeling and technical detectability serve distinct functions; neither should be conflated with the other.
Do not rely solely on a badge: Labels are helpful, but they replace neither media literacy nor journalistic scrutiny. Manipulated content can circulate without labels; conversely, a label alone does not prove that content is inaccurate.

This deliberately does not constitute legal advice. Exact obligations depend on the respective system, its role in the value chain, and the publication context. However, these five questions establish a resilient foundation before the deadline arrives.

The True Benchmark: Traceable Responsibility

The debate over AI labeling is frequently framed as a choice between innovation and regulation. That falls short. Sound transparency can make innovation socially sustainable in the first place: Those who can trace when AI was involved and who stands behind a publication can better evaluate information, address errors, and build reasoned trust.

The AI Omnibus that took effect today alters the timetable and certain administrative procedures. Yet it leaves the core untouched: Technology that transforms communication requires clear responsibility. The 2nd of August is therefore less of an endpoint than a timely occasion to make one’s own practice visible and verifiable.

Sources

Transparency and Responsibility: This article was drafted with assistance from GPT-5.6. Prior to publication, all sources were verified against primary documentation; editorial responsibility and final text verification remain with Um:bruch. Legally responsible under German press law is Lukas Geiger (V.i.S.d.P. pursuant to § 18 para. 2 MStV).

Translation: Gemini (GM). Review: Claude (CL). In case of discrepancies, the German version prevails.