AI adoption is racing ahead, but trust is not keeping up. The newest data from the Stanford AI Index 2026, Ipsos, KPMG and the University of Melbourne, Pew and Eurostat paint the same picture: most people already use AI, a majority see more benefits than drawbacks, and at the same time more than half are nervous about it. In Croatia the gap is even wider: 54% use AI tools, but only 21% trust AI systems. For companies rolling out AI to employees and customers, this gap is a project risk, not a side note.
Optimism and anxiety are rising at the same time
Ipsos has run its AI Monitor every year since 2022; the 2025 wave covered 30 countries and 23,216 adults. The Stanford AI Index 2026 summarises the trend: the share who say AI products and services have more benefits than drawbacks rose from 52% in 2022 to 59% in 2025. Over the same period, the share who say these products make them nervous rose from 39% to 52%, and two thirds (67%) expect AI to profoundly change their daily life within three to five years.
Source: Ipsos AI Monitor 2022–2025, via Stanford AI Index 2026 (fig. 9.1.1). 2025 wave: 30 countries, 23,216 adults, 21 March – 4 April 2025.
In other words, the same people hold both views. AI is seen as useful (56% expect it to cut the time it takes to get things done), but people are unsure about what it will do to jobs and the economy. Only 40% globally think AI will make their own job better; in the United States it is 33%. Europe has moved towards optimism: since 2022 the share seeing more benefits than drawbacks rose by 12 points in Germany and 10 in France, though both remain below Asia and Latin America.
Trust in AI: the KPMG and University of Melbourne study
The largest dedicated survey on trust in AI comes from the University of Melbourne and KPMG: 48,340 people in 47 countries, surveyed between November 2024 and January 2025. Its key findings:
- 66% use AI regularly and on purpose, but 54% are wary of trusting it. 72% accept AI to some degree.
- People are optimistic (68%) and worried (61%) at the same time; 73% already see benefits, 79% are concerned about risks and 43% have personally experienced negative outcomes such as inaccurate results or loss of privacy.
- Advanced economies trust AI less than emerging ones (39% vs 57%) and have less AI training (32% vs 50%).
- Only 43% think current regulation is adequate, and 87% want laws and stronger fact-checking against AI-generated misinformation. 64% worry that elections are being manipulated by AI content and bots.
Regionally, the AI Index notes that in most North American and European countries about half of employees use AI at work regularly, while trust is several points lower, between 40% and 48%. In India, China, Nigeria, the UAE and Saudi Arabia, both use and trust exceed 80%.
Who should regulate AI? Trust in governments and the EU
Asked whether they trust their own government to regulate AI responsibly, 54% of respondents agreed on average. The spread is wide: Singapore 81%, Indonesia 76%, Poland 61%, Spain 55%, Germany 49%, France 42%, Great Britain 39%, Hungary 33% and the United States 31%, the lowest of all countries surveyed.
Source: Ipsos AI Monitor 2025, via Stanford AI Index 2026 (fig. 9.3.1). EU countries highlighted. Croatia was not in the Ipsos sample.
The EU's regulatory approach is trusted more than that of the US or China. In Pew's spring 2025 survey across 25 countries, a median of 53% trust the EU to regulate AI effectively, compared with 37% for the United States and 27% for China. Within the EU, Germans and Dutch are among the most trusting of Brussels, Greeks and Italians among the least. At the same time, people do not want regulation to stop progress: globally, 58% say it is more important to advance science and medicine through AI innovation than to protect industries through regulation (41%). In Germany the ratio is 63 to 37, in France 65 to 35 and in Poland 67 to 33.
Transparency is the one point everyone agrees on: 79% say products and services that use AI should have to disclose it, a view shared in all 30 Ipsos countries. In the EU this is becoming a legal requirement. Article 50 of the EU AI Act requires that people are told when they interact with an AI system and that AI-generated content is marked. Our free Article 50 scanner checks whether a website discloses its chatbots and AI content, and our guide to AI content labelling and watermarks covers the technical side.
AI experts and the public see a different future
The gap between people who build AI and everyone else is largest in the United States. Pew asked US adults and US-based AI experts the same questions. 73% of experts expect AI to have a positive effect on how people do their jobs, versus 23% of the public, a 50-point gap. The same pattern appears for the economy (69% vs 21%) and medical care (84% vs 44%).
Source: Pew Research Center 2025 (US adults and US-based AI experts), via Stanford AI Index 2026. First three rows: share expecting a positive impact. Last row: share expecting fewer jobs.
On jobs, 64% of Americans expect AI to lead to fewer jobs over the next 20 years and only 5% expect more; among experts it is 39% fewer and 19% more. Globally, views are split: 50% think AI will create new jobs and new ways of working, but only 39% in Germany and 29% in the United States. 11% of respondents say it is very likely that AI will replace their current job within five years. More on the data in our analysis of AI and the labour market.
Croatia: usage is rising, trust is not
Croatia is not part of the Ipsos or KPMG samples, but two sources fill the gap. Eurostat measures actual use: in 2025, 27.5% of Croatians aged 16–74 used generative AI tools in the previous three months, below the EU average (32.7%) and behind Slovenia (37.6%), Austria (39.4%) and Germany (32.3%). The biggest gap is at work: only 9.0% of Croatians used generative AI for work, against 15.4% in the EU and 21.2% in Austria. Croatia is above the EU average only in education (13.4% vs 9.3%).
Source: Eurostat, isoc_ai_iaiu (individuals aged 16–74), 2025. Work = used for professional purposes; education = used for formal education.
Attitudes come from the annual national survey by Prizma CPI for Algebra Bernays University (September 2025, 2,500 adults, third wave). The share using AI tools at least occasionally rose from 35% to 54% in a year. But trust stagnated: only 21% trust AI systems and their applications, 37% distrust them. 54% have a negative first association with the term "artificial intelligence" and 40% a positive one, and the share who think risks outweigh benefits rose from 22% to 33%. Croatian AI solutions inspire less trust than foreign ones, by a ratio of about 2:1.
Source: national AI perception survey by Prizma CPI for Algebra Bernays University, September 2025, representative sample of 2,500 adults (third annual wave), as reported by Lider.
Fears centre on disinformation and work: 59% fear they will not be able to tell AI-generated content from human-made content, nearly three quarters fear AI will spread fake news, and 46% expect significant job losses within five to ten years. At the same time, 36% have no interest at all in additional AI education. That is the core problem for Croatian employers: people already use AI, but without training and with little trust.
What this means for companies introducing AI
Low trust does not stop adoption; it pushes it underground. Employees use free tools without telling anyone, customers distrust AI they did not know was there, and projects stall at the pilot stage. What works in our projects in Croatia and the DACH region:
| Finding from the surveys | What to do |
|---|---|
| 79% want AI use disclosed | Label chatbots and AI-generated content clearly, before Article 50 requires it |
| Employees use free public tools, often against policy | Provide an approved tool (e.g. Claude or ChatGPT business plans with no training on your data) and a short, clear usage policy |
| 66% rely on AI output without checking it | Define where human review is mandatory and build it into the workflow |
| Trust and literacy rise with AI training | Train teams on real tasks; Article 4 of the AI Act also asks providers and deployers to support AI literacy among staff |
| Fear of job losses (46% in Croatia) | Communicate what AI will and will not change in each role; start with tasks people dislike |
| Croatian solutions trusted less than foreign ones | Show references, data location and how the system was tested; transparency beats branding |
Limits of the data
- Online panels over-represent younger, urban and better-educated people, especially in emerging economies, which partly explains their high scores.
- Different questions, different numbers. "Trust in AI", "trust in AI at work" and "trust in government to regulate AI" are separate measures and should not be compared directly.
- Timing. The Ipsos and KPMG fieldwork took place in late 2024 and early 2025; the Croatian survey and Eurostat data are from 2025. Attitudes are moving quickly.
- Croatia is missing from the large international panels, so Croatian figures come from a national survey with its own wording.
Frequently asked questions
Do people trust AI in 2026?
Only partly. In the KPMG and University of Melbourne study (48,340 people, 47 countries), 66% use AI regularly but 54% are wary of trusting it. The Ipsos AI Monitor 2025 found that 59% see more benefits than drawbacks in AI products while 52% say these products make them nervous.
Which countries are most and least optimistic about AI?
Southeast Asia is the most optimistic: in Malaysia, Thailand, Indonesia and Singapore more than 80% expect AI to profoundly change their lives within 3–5 years. North America and Europe are more sceptical. Trust in one's own government to regulate AI ranges from 81% in Singapore to 31% in the United States, with Germany at 49% and France at 42%.
What do Croatians think about AI?
In the September 2025 Prizma CPI survey for Algebra Bernays University (2,500 adults), 54% used AI tools at least occasionally (35% a year earlier), but only 21% trusted AI systems and 37% distrusted them. 54% had a negative first association with the term, 40% a positive one. Eurostat shows 27.5% of Croatians used generative AI in 2025, below the EU average of 32.7%.
Who do people trust to regulate AI?
Across 25 countries in Pew's spring 2025 survey, a median of 53% trust the EU to regulate AI effectively, compared with 37% for the United States and 27% for China. In the KPMG study only 43% believe current regulation is adequate and 87% want laws and stronger fact-checking against AI-generated misinformation.
The bottom line
Public opinion on AI in 2026 is not for or against; it is both. A majority sees benefits, a majority is nervous, and almost everyone wants transparency. Europe trusts AI less than Asia but trusts the EU to regulate it more than the US or China. In Croatia, AI use rose from 35% to 54% in a year while trust stayed at 21%. Companies that close that gap with approved tools, training, clear labelling and human review will get further than those that only roll out licences.
Do you want AI that your employees and customers trust?
We help companies choose approved AI tools, write a practical AI usage policy, train teams on real tasks and meet the AI Act transparency and AI literacy rules. As a Claude Certified Architect from Zagreb, we work with companies in Croatia and the DACH region.
Talk to an AI consultant