By: Maida Salkanović
In October last year, millions of viewers of TV Pink—a nationally licensed broadcaster in Serbia—had a chance to watch a four-minute video in which acclaimed film director Emir Kusturica appeared to speak about his ties to intelligence agencies, admitted to “mocking his own people,” and hinted at influencing a high school shooter who killed ten people in May 2023, all while praising TV Pink’s owner. The video, however, was generated by artificial intelligence (AI).
It was a retaliatory move by Željko Mitrović, the station’s owner, following an article in which Kusturica criticized Mitrović’s influence on Serbian society. It was not the first time TV Pink weaponized fakes against dissenters. The station, closely aligned with the ruling regime, had previously produced deepfakes of prominent opposition figures Marinika Tepić and Dragan Đilas. The videos were either not labeled as AI-generated or were labeled inadequately, prompting the Association of Journalists of Serbia (UNS) to publicly urge TV Pink to stop the use of manipulative content.
Across the border in Croatia, just two months later, presidential candidate Ivana Kekin played a fake audio recording at a press conference in which Prime Minister Andrej Plenković appeared to endorse her candidacy. Unlike TV Pink, Kekin used the stunt as a public service announcement, to highlight just how easily disinformation can be manufactured and circulated. Still, the incident raises an important question: how much harder will journalists’ jobs become in a world where truth can be synthetically distorted at scale?
Artificial intelligence has become the phrase of the moment. In journalism, it inspires a range of reactions, from panic over job security to opportunistic experimentation aimed at boosting clicks and profits. The field remains largely unregulated, and amid the chaos, journalistic communities across the region are scrambling to understand these technologies, and to shape the ethical and professional standards that must govern their use.
With Little Regulation, Journalists Take the Lead on AI
Across Southeast Europe, regulatory approaches to artificial intelligence remain uneven, but momentum is building. As an EU member state, Croatia is obliged to align with the EU’s regulatory framework and, in May 2025, the Ministry of Justice, Public Administration and Digital Transition formed a working group to begin drafting the country’s national AI law to implement the EU AI Act.
In Serbia, a national Strategy for the Development of Artificial Intelligence was adopted in early 2025, and the country’s Code of Journalists, updated in late 2024, now includes references to the ethical use of AI in media.
Bosnia and Herzegovina and Montenegro currently lack comprehensive, AI-specific legislation, but both are increasingly engaging with regional and international initiatives related to AI use in media and governance. In November 2024, Montenegro signed the Council of Europe Framework Convention on Artificial Intelligence, signaling its commitment to ensuring that AI development upholds human rights, democracy, and the rule of law.
Media professionals and civil society actors across the region are increasingly engaging with the challenges and possibilities of AI in journalism, working to establish ethical guidelines and shared standards. In May 2025, representatives of press councils from Southeast Europe and Turkey gathered in Ohrid, North Macedonia, where they signed the Regional Declaration of Press Councils from Southeast Europe and Turkey on the Ethical and Transparent Use of Artificial Intelligence in the Media.
One of the first research studies on the topic was published by BIRN Serbia, with support from the OSCE, also in May 2025. The findings reveal that while journalists in Serbia are increasingly experimenting with AI tools in their reporting, they do so without adequate training, clear guidelines, or a cohesive editorial strategy.
Tanja Maksić, the author of the study, told SEE Check that her research showed journalists are looking for some form of guidance when it comes to using AI, whether through regulation, editorial frameworks, or support from self-regulatory bodies.
“The big problem here in Serbia,” she said, “is that we have a fairly authoritarian government, closed off to dialogue and any kind of consultation. It’s genuinely difficult to reach any constructive agreement with them on anything.”
Maksić sees it as a positive sign that, even within a captured media system, journalists are exploring new tools, seeking out resources, and expressing concern about the ethical implications of AI. However, she notes that copyright, which is one of the main concerns about AI globally, was rarely mentioned by those surveyed.
Many journalists are also turning their attention to how AI is affecting their peers. Slađan Tomić, a journalist working across Bosnian-Herzegovinian and Serbian media, interviewed journalists in Bosnia to better understand how they are using AI in their daily work.
“I’m afraid that AI tools like ChatGPT could, in some newsrooms, replace journalists or part of the newsroom staff, and also contribute to journalists becoming lazy and relying on the machine instead of using their own knowledge, talent, and skills. That leads to intellectual stagnation, the death of creativity, and, more importantly, the erosion of journalistic excellence. Machines cannot replace human beings, and that needs to be made clear,” said Tomić.
From Spam to Scam: Slop and Synthetic Content Flood the Web
AI, says Maksić, also generates large amounts of so-called “slop”, a term journalist John Oliver recently described as “the newest iteration of spam.” This kind of content, Maksić explains, “has no informational value, no authorial signature, clogs up platforms, obstructs the flow of information, and at times, is genuinely difficult to distinguish from real reporting.”
In most cases, AI slop is profit-driven and sometimes used in outright scams. Deepfakes of journalists are sometimes exploited to promote these schemes, relying on the trust those individuals command. In Croatia, for instance, the image of RTL editor and TV presenter Mojmira Pastorčić, alongside that of President Zoran Milanović, was used in a fabricated promotion for a supposed “investment platform,” a scam designed to extract money from citizens.
Rašid Krupalija, Editor-in-Chief of the Bosnian fact-checking platform Raskrinkavanje, told SEE Check that scammers behind these types of schemes typically choose TV journalists whose faces are easily recognizable to the average social media user. According to Krupalija, this tactic helps lend credibility to the product being promoted.
“Unfortunately, the AI tools used to create these deepfake videos are getting more sophisticated. While it was relatively easy to spot a fake in the beginning, it’s becoming increasingly difficult now, as the videos look and sound quite authentic,” he said.
Famous Bosnian journalist Senad Hadžifejzović has been a frequent target of these scams. In an effort to inform the public that the advertisements were fraudulent, Hadžifejzović aired public service announcements on his channel Face TV, urging viewers not to purchase what he called “poison and scams.” “If you want to kill yourself, then buy it,” he said, referring to the warnings broadcast on the station, though the impact of those warnings remains unknown.
He told SEE Check that they reported the matter to the Prosecutor’s Office but never received a response. “It’s all up to the investigative bodies and prosecutors: we don’t know that they’ve done anything,” he said.
Some social media platforms now require AI-generated content to be labeled, but the approach remains deeply flawed. Meta, for example, mandates labeling only for “photorealistic video or realistic-sounding audio” created using AI, excluding images from this requirement. Krupalija warns that AI-generated content will inevitably become more difficult to detect, a development that works in favor of those using it for scams.
“It is the job of journalists to investigate and report on these types of fraud,” Krupalija told SEE Check, “but reporting alone won’t be enough without efforts to promote media and information literacy among readers. That is, building their resilience and skills to recognize scams and other harmful online content.”
Such misuse not only defrauds the public but can erode trust in the very journalists whose identities are being co-opted. Over time, this could contribute to a broader crisis of confidence in the media profession itself, especially in regions where trust in journalism is already precarious. A notable example is the Slovak case, where a deepfake audio clip mimicking journalist Monika Tódová was used to fabricate a conversation about election fraud. The aim was not only to discredit the political candidate, but also the journalist, showing how AI tools can be weaponized to undermine both individual credibility and public trust in journalism as an institution.
The phenomenon underscores a key vulnerability that extends beyond scams: the increasing difficulty in detecting AI-generated content poses a serious threat not just to consumer protection, but to the integrity of political reporting and journalism more broadly. As synthetic content becomes more convincing, the risk of its abuse in politically motivated smear campaigns, disinformation efforts, or fabricated interviews grows. In this landscape, identifying AI-generated content is not only a matter of countering fraud: it is critical to preserving editorial standards and democratic accountability.
AI as an Assistant: Supporting Journalists, Not Replacing Them
AI can undoubtedly be a valuable tool for journalists, enhancing both productivity and efficiency. This potential was recognized by the Journalists’ Association of Serbia (UNS), which partnered with the Center for Youth Activism Development (CROA) from Sarajevo, the Albanian Center for Quality Journalism (ACQJ) from Tirana, and the Media Center from Čaglavica to implement the project “How to Use Artificial Intelligence in Journalism.” The initiative resulted in a regional research study on the use of AI in media, as well as the creation of a regional network through which young people exchange experiences, tools, and training related to AI applications.
Aleksandra Ničić, one of the trainers in the project, said she was surprised by the level of knowledge demonstrated by young and aspiring journalists, which was higher than expected across the board. She also noted subtle regional differences in focus and concerns.
“My impression is that participants from Belgrade were primarily interested in the practical use of AI tools, those from Sarajevo were more focused on the privacy implications, participants from Čaglavica engaged most with the issues of AI hallucinations and deepfakes, while participants from Tirana were the most active in discussing the broader consequences of AI in journalism,” she said.
Ničić believes that there will be no serious negative consequences for the profession, as long as artificial intelligence is treated as an assistant that supports journalistic work, rather than a colleague doing all the work and creating problems through hallucinations.
“In other words, if we use AI tools to handle repetitive or mechanical tasks that consume our time, or to generate ideas that we then critically assess, I believe there won’t be negative consequences for our work,” she said.
Journalist Slađan Tomić agrees that AI can be a valuable tool when used as an assistant, but not as a content generator. He noted that there are already examples of media outlets using AI to produce content with no human oversight. In some cases, he added, journalists fail to even remove the AI’s built-in prompts, such as ChatGPT’s phrase: ‘If you want a shorter version, I can write one,’ which ends up appearing in the published article.
The use of AI appears to be more advanced in media outlets with lower editorial standards. In Serbia, Maksić observed that tabloids are among those furthest along in integrating AI tools into their content. These outlets frequently rely on text-to-speech and text summarization tools, often with poor results.
“It’s a mechanical voice, it gets the cases wrong, the diction is off, you can clearly tell,” she said. “But they’re using it consistently, and at some point, they’ll probably bring it to an acceptable level.”
According to Maksić, the combination of lower editorial standards and greater financial resources allows these outlets to experiment more freely with AI technologies than their peers in more rigorous journalistic environments.
“When you consider that these outlets already have a tendency to produce propaganda, disinformation, and fake news, and that they’ve been heavily engaged in smear campaigns against independent and critical voices in society,it becomes genuinely worrying,” said Maksić. “The issue now is that AI will enable them to further multiply low-quality content. They’ve already shown they’re willing to do it without AI, and now they’re gaining a powerful new tool to spread and accelerate these harmful practices.”
New Tech, Old Questions
AI continues to dominate conversations in Southeast Europe. New regulations are being drafted, media trainings are underway, and professional circles are actively debating how to approach this fast-evolving technology.
But this isn’t entirely new terrain, Maksić points.
“We started with printed newspapers, then came social media, and now AI. Over time, the profession has shown a certain resilience, and an ability to integrate new technologies into its workflows,” she said.
There are many new challenges ahead, but many journalists believe they can be overcome, so long as AI is approached with caution, responsibility, and a strong ethical foundation.
“A machine can be a tool and a support to the human, but nothing can replace human thought,” Tomić concludes. “Through irrational, unethical, and excessive use of AI in journalism, we risk permanently devaluing the profession.”
