How can a complex area of news such as EU affairs be made more accessible and easier to understand? This is the question that 15 legal entities representing around 40 media outlets across the European Union wondered.

And they found the creation of a completely new and innovative chatbot called ChatEurope as the solution. The main idea is to bring together trusted content from the different European media partners to offer quick summaries to users in addition to allowing them to navigate European affairs more easily.

Led by the Agence France-Presse (AFP), the media consortium behind this project included the German Deutsche Welle, the French France Médias Monde, the Romanian radio RFI Romania, the German and Italian news agencies Deutsche Presse-Agentur (DPA) and Agenzia Nazionale Stampa Associata (ANSA), the Polish press group Agora, the Spanish fact-checking organisation Maldita, the Spanish daily El Pais, and the online media outlet specialised in Southeast Europe OBCT.

Co-funded by the European Commission but completely independent, the project, which is still in its testing phase, hopes to become increasingly well-known.

The Fix spoke with Mirko Lorenz, innovation manager at Deutsche Welle.

How did the idea for this project come about?

The idea was mainly to find a way to help European citizens understand EU affairs better. It’s now quite a large topic because it affects all kinds of things, politics that you don't care about, and specific new regulations that directly affect your European life.

The EU had several calls last year where they basically wanted to fund efforts to help people understand the EU better. So we asked for funding, with the AFP as the main driver for us, with quite a large group of European media, 15 partners, which represent 40 different media outlets.

What we are trying to do is create a big database of European news that can be trusted from official sources like AFP, Deutsche Welle, DPA, or ANZA. And the core innovation is that this database is now accessible not through a search, but through a chatbot, an AI-supported chatbot.

So this generative AI-powered chatbot is central to this project?

It allows users to ask questions and receive tailored answers in their preferred language, with full transparency regarding sources and links to original content. What is very important to note as of this moment is that this is an early experiment, meaning that the database is limited at this time. We need some time to build it up.

You can ask, for example, “I’m from Poland, and I would like to work in Spain. What do I have to consider to actually move to Spain?” And the platform will then give you a very careful answer and base all the answers and all the facts on actual sources.

That is a low-level example of what we are after to provide accurate news about not just this topic. You can also ask more complex current political things, like what is the position of the EU on Gaza or what is the EU doing to support Ukraine? And that's where the value could lie over time.

What about the translation of the articles available on the platform? Is it done automatically?

The project plans to produce approximately 2,700 diverse digital news items, including 1,064 exclusive original productions and 1,670 curated pieces, available in 7 native European languages, with advanced translation and transcription tools making content accessible in all EU languages.

Initially we used DeepL, which is a really good platform for translations, but now we are implementing plain X, a self-developed tool out of research projects together with a Portuguese partner of Deutsche Welle. One of the benefits of plain X is that we are a hub for AI engines, meaning that we can provide transcription, translation, and even subtitling for videos or voiceovers.

And it's semi-automated. It can be fully automated, but there is the option for the human in the loop, meaning that an editor can actually go in there. In this way, we want to first service seven core European languages, but if the technology allows, we will extend coverage to all 24 European languages.

And the big hope for the future is that we can even cover local dialects. In plain X, we are currently going very much after low-resource languages. Low-resource is not about the money that's available for a language but the training material.

Is this project also a way to fight misinformation?

Actually, it's interesting to use AI, but you need to be very, very careful that you're not opening the door for falsifications. So ideally, we have a really large database, which you can interrogate using AI, but get out trustable summarisation.

The big chance of AI combined with news databases is that the search is better. Very often it was better to go to Google to search for the news instead of for the search on a local site. So with AI, it creates a summary, but based on the words in the article. It's very important to note that this is not like generative AI that would come up with the risk of hallucinations.

Added to that, the system that we have set up is a RAC, retrieval augmented generation. It means that our AI is not allowed to go to the Internet to search for answers. It is restricted in the answers it can give.

An important point is that you received money from the European Union, but you’re totally independent?

Yes. [EU funding is]It's just an enablement to do this technical project, but there's no influence whatsoever in what news is reported. It can be critical of the EU as well. Nobody would interfere with that.

Do you think this can be the first step for a better collaboration between different media outlets in Europe?

At least it's a very important testing ground, because relevant European media is there, and combined with this exciting new technology, which is AI.

I am happy that I find a lot of motivated colleagues from all over Europe working together in this case. You would kind of expect more competition, but that's not the case. Maybe, just maybe, there's a certain understanding that with the pressures from outside, we need to create a European public network.

What does the future hold?

We will also adapt to modern news consumption habits by producing social media-friendly formats like reels and short videos, aiming for a broad reach among EU citizens of all ages.

The project is slated to run for 24 months. I think these types of projects can be prolonged if they are successful. But this is not decided as of this moment.

Source of the cover photo: kuu akura via Unsplash

Correction: an earlier version of this article contained a spelling error in the name of the plain X tool.

ChatEurope

Meet ChatEurope, the new innovative chatbot for European news

Most popular

Meet ChatEurope, the new innovative chatbot for European news

The European Correspondent’s quest to build a Europe-wide media outlet

How Aftenposten reinvents its homepage with AI to increase engagement and subscriptions

How publishers can master Google Discover – practical tips for news leaders

FixEd podcast: Aleksandar Manasiev on mastering multi-platform journalism

Canadian publishers are still adapting two years after Meta's news ban

Subscribe to The Fix's newsletter courses