On Radio Sljeme: what Croatians fear about AI and where it is heading
Tomislav Vazdar, founder and CEO of Riskoria, joined Croatian station Radio Sljeme for a programme about artificial intelligence.
It covered what Croatians think of it, how justified their fears are, and where its development is heading next.
The starting point was a CRO Demoskop survey. The agency Promocija plus carried it out for RTL in February 2026 among 1,000 respondents. Vazdar summed up his view of the whole subject in a single line: “Artificial intelligence hasn’t invented a new kind of thief. It has given the old one a better tool. What we need isn’t fear, but a few new habits.”
What Croatians think about AI
For years the public debate has revolved around robots taking jobs, but Croatians see it differently. The danger they named most often was inaccurate and fake information, images and recordings (29.5%). Next came misuse of data and privacy (26.4%) and excessive control over people (20.7%). Job loss came only fourth, at 19.3%. Vazdar sees an accurate judgement here rather than panic: “People aren’t afraid of the technology. They’re afraid of not knowing who is in control of it.”
The largest group, 38.7% of respondents, said AI is equally useful and dangerous. “That isn’t fear. It’s people waiting for someone to explain it honestly,” Vazdar said. Some 59.3% first heard about AI from the media, and only 4.7% at school or university. The media therefore shape what most people think, and sensational headlines about machines breaking free push the undecided towards fear. Reassurance does not cure that fear, Vazdar argued; knowing what to do does. Once someone has a plan, fear turns into caution, and caution is healthy.
A fear with real grounds
Worrying about fake recordings is not an overreaction. A scammer anywhere in the world now writes flawless Croatian. He can build a voice that sounds like your child from a short clip on social media. He messages thousands of people at once. “The thief hasn’t changed. The cost of his work has.” The European Union Agency for Cybersecurity says much the same in its ENISA Threat Landscape 2026. Attackers primarily use consumer-grade AI tools to augment the skills they already have.
So the answer is not learning to spot a fake, which keeps getting harder. “A voice is no longer proof. Neither is a picture. The only proof is a call you made yourself, to a number you already have.” Every scam shares the same skeleton of urgency, secrecy and money. Vazdar reduced the defence to three words, stop, hang up, verify. He added one measure that costs nothing, a family password only the household knows.
Where AI is heading
Much of the conversation looked ahead, and Vazdar separated three ideas that are often blurred in public. Today’s artificial intelligence was not written rule by rule. It learned particular tasks from vast numbers of examples, and at those tasks it can outperform people. Artificial general intelligence would be a system able to learn any intellectual task a human can, and it does not exist today. Superintelligence, a system smarter than all of us at everything, is for now only a hypothesis. “Every scam we’re talking about today is done with the first, ordinary kind,” he said.
The real shift on the way is a different one. Until now we asked AI a question and got an answer. The new generation of systems is given a task and carries it out. It looks, works out a plan, acts, checks whether it worked and tries again. Such agents will soon be buying, booking and paying on our behalf. “It’s the difference between an adviser and someone who actually does the job,” Vazdar explained. Useful, but it means we will soon be protecting not just ourselves but our digital assistant as well.
Scams will also become more personal, because the message will know where you work and where you spent your holiday. We published all of it ourselves. At the same time, the same technology already helps banks spot suspicious payments and stop them. That makes this a race rather than a lost battle. On the fear of machines escaping human control, Vazdar stayed grounded. What has happened so far are not uprisings but human mistakes, where someone gave a tool a task and too much freedom. “The tools will change every year. The trick has been the same for centuries.”
What it means for organisations
The survey shows that Croatians use AI tools far more at home than at work. That means employees are discovering them on their own and without rules. The EU Artificial Intelligence Act already expects organisations that deploy such systems to support AI literacy among their staff. With systems that act on their own, this becomes a security question too. An AI agent should get only the permissions it needs, and someone should be able to switch it off within a minute.
More than a third of Croatians have not yet made up their minds about AI. Whether they move towards fear or towards caution depends on who explains it to them first, and how honestly. It is worth asking whether your own staff would know. Riskoria helps organisations with staff awareness and the secure adoption of AI.
Tomislav Vazdar leads Riskoria, an advisory firm for cyber security, risk management, artificial intelligence and data protection. He has worked in the field for more than twenty five years and speaks at conferences and to the media.
Related reading
Want an opinion on your own situation, not a general text?
Half an hour, no obligation. Instead of a general text you get an answer for your organisation.

