Why did AI become experts’ top future risk?
In AXA’s 2026 perception survey, AI and big data rose from fourth to first among 4,345 risk professionals, while 24,000 adults ranked it second behind climate change. The result measures concern, not the probability or cost of future harm—and 89% of the expert panel worked for AXA.
Editorial responsibility: The Impact of AI Editorial Desk · Report a factual concern
At a glance
- 1AI and big data ranked first among 4,345 expert respondents and second among 24,000 adults; climate change ranked second for experts and first for the public.
- 2Respondents selected and ordered five of 20 risks by expected impact over the next five to ten years, so the ranking measures comparative perception rather than probability, expected loss or observed harm.
- 3The expert sample is commercially and institutionally concentrated: 89% worked for AXA and 11% came mainly from its professional networks.
Research topic
How experts and nationally representative adult samples ranked AI and big-data risks in AXA's 2026 cross-country perception survey
The answer: concern rose sharply, but this is a perception ranking
Artificial intelligence and big data moved from fourth place in 2025 to first in the expert ranking of AXA's 2026 Future Risks Report. Among the general population, the category rose from seventh to second, behind climate change. That is a meaningful change in what respondents regard as consequential over the next five to ten years. It is not an estimate that AI harm is now more likely or more costly than climate, cyber or geopolitical risks.
The distinction is essential because the survey asks people to rank concerns. Views can react to product launches, media coverage, labour-market anxiety and recent incidents even when underlying probabilities are unknown. The report captures the social and professional salience of AI at a particular moment. Policymakers and employers can use that signal to identify trust and preparedness gaps, but not as a substitute for incident data, exposure measurement or cost-benefit analysis.[1][2]
The Impact Brief · Free
Follow the evidence in society & media.
Get a five-minute weekday briefing on what changed, why it matters and where the evidence comes from. Choose the topics you care about.
What Ipsos asked and how the ranking was built
Ipsos ran online questionnaires for AXA from 14 May to 24 June 2026. Respondents saw 20 risks and selected the five they believed could have the greatest impact over the next five to ten years. A first-ranked choice received five points, the second four and so on. The published order is therefore a weighted preference ranking, not simply the percentage of people who mentioned each risk.
The study included 4,345 expert respondents across 62 countries and regions. It also surveyed 24,000 adults aged 18 or older in 19 countries and regions, using nationally representative samples of 1,000 in most markets, 2,000 in Germany and 5,000 in the United States. The report does not publish sampling weights or margins of error, limiting how precisely readers can compare close positions or small differences between countries.[2]
The expert label needs an important qualification
Eighty-nine per cent of the expert panel worked at AXA, mainly in distribution, marketing and risk management. The remaining 11% came from AXA's professional networks, mainly in financial services. These respondents have relevant exposure to risk work, but they are not an independent probability sample of scientists, regulators, labour economists, technology auditors or the wider insurance industry.
That concentration may shape which risks feel urgent and how categories are understood. AXA commissioned the research and uses it to inform strategy, risk management and public positioning. The report also presents insurance products, prevention initiatives and resilience investments. None of that invalidates the responses, but it means the expert result should be described as the view of an AXA-centred professional panel rather than a global expert consensus.[2]
What respondents said worries them about AI
Among people who selected AI and big data as a leading risk, 54% of experts and 48% of adults described it as rapidly emerging. Sixty-five per cent of those experts and 67% of those public respondents said they felt vulnerable to it in daily life. Ethical concerns—including human rights, data misuse and bias—were the leading specific concern for 34% of each group. Effects on jobs followed at 28% among experts and 30% among adults, while lack of transparency or the 'black box' problem drew 21% and 17% respectively.
These percentages are conditional on having selected the category, not shares of the entire expert and public samples. The report should therefore not be paraphrased as saying two-thirds of everyone surveyed feels personally vulnerable to AI. The figures help explain the ranking: respondents connect AI with rights, employment and opaque decisions rather than treating it as an abstract technical issue.[2]
Preparedness looks weakest where governance responsibility is clearest
Only 9% of experts who assessed AI-related risk said public authorities were well prepared; in Europe the figure was 3%. Among the public respondents who chose the risk, 29% judged authorities well prepared. The gap may reflect professional respondents' closer view of regulatory and operational complexity, but the survey cannot determine why the groups differ or independently test actual government capacity.
For public-service professionals, the practical implication is not to chase reassurance through slogans. Agencies can make readiness visible through incident reporting, appeal routes, procurement records, named accountability and tests that involve affected workers and service users. Employers can address job concerns by explaining which tasks are changing, what review remains human and how staff can challenge harmful outputs. Transparent process is more actionable than trying to persuade people that a high-level risk category is simply over- or under-rated.[2]
Geography complicates the global headline
AI and big data ranked first among experts in every broad region except Europe, where geopolitical instability led. Climate change remained in the top three everywhere except Africa. These differences show why the overall table should not erase local context: conflict exposure, economic structure, demographic pressure, recent disasters and access to technology influence which future threat appears most immediate.
Year-on-year comparisons also require care. Colombia and Thailand entered the public sample while Switzerland left, and the questionnaire changed some categories. Ethical technology risk was merged into AI and big data, four categories were removed, two pollution-related categories were combined and social fragmentation was added. Part of the ranking shift may therefore reflect a changed menu as well as changed concern.[2]
What this means for workers, institutions and insurers
For workers, the report shows that employment consequences sit close to the centre of public concern. It does not establish the number of jobs that will be lost or created. Employers should pair adoption plans with task-level workforce evidence, training, consultation and monitoring for workload transfer. For teachers, clinicians and other professionals subject to automated decisions, rights, data use and explainability need to be part of implementation rather than after-the-fact communications.
For insurers and financial firms, the result supports closer monitoring of model failures, cyber dependencies, liability and business interruption. It does not prove a particular product is necessary or appropriately priced. Because AXA has a commercial interest in risk prevention and insurance solutions, readers should separate the survey's reported responses from the company's strategic recommendations.[2]
What would change the assessment
Confidence in the trend would rise if independent surveys using stable categories and transparent weights reproduced the same shift, and if longitudinal panels showed that individuals—not just yearly samples—changed their views. Measured incidents, losses, discrimination, job transitions and service disruptions would help connect perception to observed impact. Publication of uncertainty intervals would also clarify whether neighbouring ranks are meaningfully different.
The report itself warns that perception surveys should be read in light of events during fieldwork. The fair conclusion is therefore neither that experts have proved AI is the world's largest risk nor that concern is merely hype. A large, geographically broad survey shows AI has become a leading perceived threat, especially around ethics, jobs and opaque decisions. The next evidential step is to test those concerns against independent, outcome-based measures.[2]
What this means for people
- Workers' job concerns are visible in the survey, but the ranking cannot predict how many roles or tasks will change.
- People want clearer accountability for rights, data use and opaque automated decisions; institutions can respond with concrete appeal, audit and incident processes.
- Public authorities should not treat a low preparedness rating as a performance measure, but it is a warning that governance capability is not visible or trusted.
Global context
The survey spans 62 countries and regions for its AXA-centred expert panel and 19 for nationally representative public samples. Regional rankings differ, and the public-country mix changed from 2025. The result is globally broad but not a single universal risk judgment; economic conditions, conflict, climate exposure and digital adoption shape local priorities.
What the evidence does not yet show
- This is a perception survey, not an estimate of risk probability, expected financial loss or observed harm.
- Eighty-nine per cent of expert respondents worked for AXA, and the remainder came mainly from its professional networks.
- The report does not publish sampling weights or margins of error for the comparisons used here.
- Question categories and the public-country sample changed from 2025, complicating year-on-year comparisons.
- Several AI percentages apply only to respondents who selected AI and big data as a leading risk, not to the full samples.
- AXA commissioned the study and has a commercial and strategic interest in risk prevention and insurance services.
What to watch next
- Independent surveys using stable categories, transparent weighting and confidence intervals
- Observed AI incidents, losses, discrimination cases and job transitions by sector and country
- Whether public-authority preparedness improves in repeat surveys after concrete governance measures
- Regional evidence that separates technology concern from conflict, climate and economic conditions
Living evidence record
Impact record IAI-0FW26OZ
Evidence stage
Observed
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Not yet
Record status
Monitoring
Last checked
13 October 2026
Source trail
2 direct sources across 1 source type.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Related-source reporting disclosure
This record analyses 2 linked source records around the same underlying development. The extra records add method, date or context, but they do not by themselves constitute independent replication of every performance claim or predicted outcome.
Evidence trail
Sources used for this report
Links checked 13 October 2026
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Continue the story
Related reporting
Society & Media
How did fake experts reach real newsrooms?
OpenAI says two covert influence operations used false journalist identities and a front research centre to place material in real outlets. AI mostly helped with drafting, translation and internal reporting; the harder failure was identity and source verification, and claimed reach remains only partly corroborated.
10 min · 2 sources
Society & Media
Do safer-driving AVs change human behaviour?
A peer-reviewed quasi-experiment found that Lyft automated vehicles followed more smoothly and conservatively, while human drivers left shorter gaps behind matched automated leaders. The study analysed real trajectories, but it did not randomise drivers or measure crashes, injuries or current vehicle systems.
8 min · 3 sources
Society & Media
What do people actually ask image-upload AI to do?
An unreviewed Microsoft-led study analysed 42,617 de-identified Copilot image-upload sessions and checked its taxonomy against 23,413 ChatGPT sessions. Users often chained recognition into writing, code and data tasks—but private, model-generated summaries limit what outsiders can verify.
8 min · 2 sources
Reader commentary
Add evidence, experience or a question
No account is required. Reader notes are published after a brief civility, relevance and safety check; disagreement is welcome.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.