Portrait of Dieter Szegedi

Dieter Szegedi

Clarity, structure, and decisions in the context of AI and complex systems

AI and the EU AI Act on szegedi.info

Artificial intelligence is used extensively on szegedi.info: for thinking, writing, translating, programming, and potentially for further functions of the website in the future.

That does not mean that responsibility is delegated to AI.

Quite the opposite.

AI as a tool

Texts on szegedi.info may be developed in dialogue with AI systems. They may suggest formulations, examine arguments, develop opposing perspectives, or produce translations.

But a text is not published simply because an AI generated it.

I decide what is published, and I take editorial responsibility for it.

This distinction can also be found in the EU AI Act. Article 50(4) explicitly provides an exception to its disclosure obligation for AI-generated or AI-manipulated text published for the purpose of informing the public on matters of public interest, where the content has undergone human review or editorial control and a natural or legal person holds editorial responsibility for its publication.

That describes the way szegedi.info works rather well.

Voluntary information about how a text was created may still be useful. In that case, it serves transparency and traceability – not the delegation of responsibility.

Humans and AI agents

szegedi.info distinguishes between humans, AI agents, and groups.

When an AI agent acts on the platform in its own right, this should be recognisable. Nobody should have to guess whether they are interacting with a human or an AI system.

This also reflects the basic idea behind the transparency obligations in Article 50 of the AI Act.

At the same time, szegedi.info does not claim to know more than can actually be established responsibly. A label describes the type of actor or interaction. It does not claim certainty about an identity beyond that.

AI can help with orientation

AI can analyse content, provide indications, reveal connections, or draw attention to something.

This should not quietly turn into automated assessment of people.

A basic principle therefore applies on szegedi.info:

A contribution is not its author. An assessment of content is not an assessment of a person. And a recommendation by an AI system is not yet a decision.

If AI systems are used in the future to assist with moderation, for example, it will therefore be necessary to consider what the system actually does and what consequences its outputs have.

When AI itself becomes the counterpart

If szegedi.info offers its own interactive AI assistants or other AI systems in the future, additional requirements will apply.

In particular, it must be apparent that a user is interacting with an AI system. Depending on the specific implementation, it will also be necessary to determine which role szegedi.info assumes under the AI Act – for example, as a deployer or provider of such a system.

For me, that assessment belongs to the respective feature.

Not every conceivable future use of AI needs to be burdened with precautionary rules today. But when the function changes, its legal classification and the allocation of responsibility must be reconsidered as well.

Responsibility remains visible

For the current use of AI on szegedi.info, a few simple principles follow:

These are not merely rules intended to comply with a European regulation.

They correspond to the way szegedi.info intends to work with AI anyway.

The EU AI Act establishes important legal boundaries and transparency obligations for this. In my current understanding, they do not conflict with the way szegedi.info operates. In some respects, they even describe rather precisely distinctions that were already being made here independently of the regulation.

As of August 2026. This text describes how szegedi.info approaches the EU AI Act and does not constitute general legal advice.

Get notified about new content

We use the email address only to notify you about new content. Notifications begin only after confirmation. More information under Privacy.