Artificial intelligence (AI) is no longer only knocking on newsroom doors — it is actively used in daily journalism workflows. From summarizing short news items and creating videos from articles to assisting with research and editing, AI tools are embedded in practice. The debate has therefore shifted from whether AI will arrive to how journalists and newsrooms will coexist with it.
Door Tamás, a certified business AI expert and head of Mediaworks' artificial intelligence project, said on the podcast A nagy AI-sztori: "It is not AI that takes a journalist's job away, but the journalist who knows how to use AI." The podcast's regular expert guest is Aczél Petra, a communication researcher and professor at Széchenyi István University.
What AI is good for in journalism today
One of the clearest strengths of generative models is rapid, structured processing. According to Door Tamás, they are particularly efficient for short news items, international summaries, and daily economic or political roundups: they can compile the important news by morning, help organize data, and free up journalists' time. Where the purpose of content is merely quick information transfer, language models perform very well.
However, the discussion underlined that this does not imply the whole profession can be automated: for longer, magazine-style or narrative pieces, human uniqueness, authorial voice, and critical thinking remain decisive. Aczél Petra pointed out that the internet had already begun to erode traditional notions of authorship — online, it is less clear who a voice belongs to — but those who retain a distinct personal voice still stand out. Longer, deeper works are increasingly seen as "craft" content: who wrote a piece matters as much as what is written.
Video and other formats — where is the boundary?
The podcast also covered that AI can now produce video and apply lip sync techniques, raising new questions about authenticity and presentation. Such technologies add further challenges for newsrooms, especially when they influence the final form and visuals of a piece.
Regulation and data security: legal frameworks lag behind
While models change rapidly, legal frameworks are only now trying to catch up. The European Union's AI Act makes some form of labeling for AI-generated content mandatory from August 2, but in practice there is still no clear, detailed guidance on exactly what must be disclosed. The discussion raised specific questions: should a journalist who used ChatGPT to prepare for an interview disclose that; or how should a video be classified if only the lip sync is AI-generated but the person is real?
Currently, regulation provides more of a conceptual framework than concrete answers, so newsrooms need to create internal policies.
Another crucial issue is data security: what do journalists upload to models and what happens to that information? Does the system really not learn from it? Aczél Petra argued this is not only a technological issue but a cultural one: organizations must define clear internal rules about what can and cannot be uploaded into AI systems. Door Tamás put it simply: today we communicate with Chinese and American servers while often not knowing exactly what happens to our data.
Practical guidance and further podcast content
The podcast walks through how AI changes journalists' day-to-day work, which kinds of content AI already produces more efficiently, where authorial voice and critical thinking persist, AI's role in video production, and how newsrooms should handle freed-up time and AI strategy. Several topics were referenced with timestamps (for example: How does AI change journalists' daily work? – 01:16; AI content labeling will be mandatory — but how? – 18:24).
(Photo: Arda Savasciogullari)
