The EU AI Act is the European Union’s regulation on artificial intelligence, which establishes a risk-based framework for the development and use of AI systems. The regulation sorts applications into different risk categories and imposes transparency obligations on content production. For public relations and communications professionals, it sets out the legal and ethical conditions for publishing AI-generated content. The key concept is editorial responsibility. It is important to understand how the EU AI Act affects PR, as the legislation fundamentally changes content management processes.
Key takeaways
- The EU AI Act establishes a risk-based framework for the development and use of AI systems.
- Communications professionals must comply with new transparency and labelling obligations in content production.
- Editorial responsibility and human review can exempt certain text content from mandatory labelling.
- For audiovisual content and deepfakes, labelling is mandatory in every case to ensure authenticity.
- PR contracts and internal processes should be updated to meet the new legal requirements.
The EU AI Act and the 2026 deadlines
Contrary to popular belief, the EU AI Act didn’t come into force just now, but on 1 August 2024. The confusion stems from the fact that its various elements have been becoming applicable in stages ever since, and on 2 August 2026 most of the regulation kicked in, including the transparency rules that affect the communications profession most. Meanwhile, however, this July the EU pushed back certain deadlines in the regulation (but not all of them!). That muddies the waters a little too, but more on that later. First, a quick detour into definitions.
The AI Act’s risk categories and the roles of communications professionals
In short, the AI Act doesn’t ban the technology; it sorts the ways it is used into categories. These are:
- Minimal risk
- Subject to transparency obligations
- High risk
- Prohibited
The AI Act also defines important roles: the provider and the deployer. The former are the developers of AI systems (think OpenAI, Google or, indeed, Anthropic). Deployers are those who use AI systems for various professional purposes – like us, PR and communications professionals (unless, of course, we happen to be sitting in the head of communications chair at Anthropic – but let’s not get lost in the details…). We need to keep analysing how the EU AI Act affects PR work in our day-to-day operations.
The bottom line: the good news for us is that the toughest rules apply to providers. Still, we can’t just sit back, because we deployers have obligations too. These include the labelling and transparency obligation, building AI literacy (yes, there is such a rule, although it has been somewhat amended since the original version) and steering clear of prohibited practices. There is also a set of rules for high-risk systems. And this is where we need to circle back briefly to the question of what took effect in August 2026, because the introduction of this part was delayed by more than a year – another source of misunderstanding. Below, let’s look at what the labelling and transparency rule means in practice.
Labelling rules for AI-generated text and the editorial responsibility exemptions
As I mentioned in the introduction, one of the basic misconceptions concerns the general, across-the-board labelling of texts produced with AI. The rule, however, says that AI-generated texts that inform the public on matters of public interest (e.g. news summaries, information notices, etc.) must be labelled, unless they have been reviewed by a human and someone takes editorial responsibility for them.
This is a very important point, because it contains two “loopholes”. On the one hand, not every text used in corporate or marketing communications counts as informing the public on matters of public interest. On the other, even for public-interest texts, meaningful editorial review and human accountability provide an exemption from the labelling obligation.
It’s worth noting, though, that there is no regulatory practice as yet. A big question is to what extent a press release, a corporate blog post or a LinkedIn post counts as public-interest information. And what counts as meaningful editorial review may also need clarifying. The boundaries of the labelling obligation are currently about where the State of Michigan’s were in the 1830s: still being drawn.
Transparency obligations for audiovisual content and deepfakes
Unlike text, the rules for audiovisual content generated or modified with AI are crystal clear. There is no exemption for editorial approval and no appeal to content not being of public interest: deepfakes must be visibly labelled. According to the EU AI Act’s definition, a deepfake is image, audio or video content generated or manipulated with artificial intelligence that resembles existing persons, objects, places or events and could appear authentic, but isn’t (as opposed to the everyday sense, where a deepfake is deceptive content about people, made with some kind of malicious intent). In practice, this means that an AI-generated influencer, spokesperson or model is just as much a deepfake as a realistic-looking location or event photo, or even a voice clone in a radio spot. There is one small concession here, though: for creative, satirical or fictional content, the labelling can be discreet.
The impact of the EU AI Act on client–agency collaboration and PR contracts
The practical application of the EU AI Act adds new elements to client–agency collaboration as well. Without claiming to be exhaustive, let me raise a few. For text content, it’s definitely worth defining at the start of the collaboration who counts as the “responsible editor”. Signing off on finished materials has always been a key task, but this role now has an added dimension. It may be worth setting this down in the contract when defining the approval process. The same approach should apply to defining labelling practices and responsibilities for audiovisual content. And that’s before we even get to the tools used by the various video and photo subcontractors. My advice: everyone should update their company contract templates, because I have a feeling that sooner or later, handling this topic carelessly could land companies with hefty fines.
Compliance checklist for heads of communications
The new regulation is likely to raise a whole host of questions among professionals. Below, I’ve gathered a few topics worth thinking through in relation to our day-to-day work.
- Do we know which AI tools our team and our partners use?
- Who approves published content in our organisation, and is this traceable?
- Do we have internal rules on AI-generated texts, images and videos?
- What properly documented steps have we taken to build our team’s AI literacy?
- Do we have any AI tools in our company that might fall into the high-risk category?
Answering these questions is an important step towards meeting the requirements of the EU AI Act.
Summary: the role of credibility and human accountability in the age of AI
All things considered, I believe the EU AI Act essentially states that credibility in the age of AI (too) depends on human accountability. This is not a new challenge for communications professionals, but one of our profession’s most important core values. We have always been responsible for the credibility of the information we publish. Now this fundamentally ethical question has also taken on greater legal weight. If you’d like to dig deeper into this topic, feel free to get in touch with us at rnagy@probako.hu. I also recommend the European Commission’s AI Act Service Desk, which offers the full text of the regulation and explanatory resources in English, as well as the Commission’s page on the Code of Practice on AI-generated content, which explains how labelling works in practice. For national enforcement, it’s worth checking the website of the AI market surveillance authority in your own country.
Frequently asked questions
- How does the EU AI Act affect PR in day-to-day work?
The EU AI Act imposes new transparency and labelling obligations on PR professionals in content production. The regulation emphasises the role of editorial responsibility, which allows an exemption from mandatory labelling where human review takes place. To comply, professionals need to document their use of AI tools and their teams’ AI literacy.
- When is labelling AI-generated text mandatory?
AI-generated texts must be labelled if they inform the public on matters of public interest. The regulation makes an exception if the content has undergone human editorial review and a person takes responsibility for it. In corporate and marketing communications, the boundaries of the labelling obligation will depend on how regulatory practice develops.
- What is the difference between the rules for text and audiovisual content?
For audiovisual content and deepfakes, the EU AI Act sets stricter rules than for text. There is no exemption through editorial approval: deepfakes must be visibly labelled. This applies to AI-generated influencers, spokespeople and any realistic-looking but artificially produced image or audio material, with the exception of creative works, where labelling can be discreet.
- Why is editorial responsibility important under the new AI regulation?
Editorial responsibility is a central element of the EU AI Act, safeguarding credibility in the age of AI. This legal framework reinforces the core ethical values of communications professionals, as human review now carries legal weight. Together, transparency and accountability define the new compliance requirements for PR contracts and client–agency collaboration.
