In a University of Melbourne study run with KPMG, Canadians were more cautious about artificial intelligence than people in most of the 47 countries surveyed. Fieldwork ran from November 2024 to mid-January 2025. Only 34 percent of Canadian respondents said they were willing to trust AI, 70 percent expected it to deliver benefits, and 79 percent were concerned about negative outcomes.
Key takeaways
- The survey is a dated snapshot of opinion, collected online from 48,340 people, with roughly 1,000 respondents per country.
- In Canada, 46 percent said AI's risks outweigh its benefits. The global figure was 32 percent.
- Canadian respondents reported the largest benefit as less time on routine tasks and the largest risk as less human interaction.
- Only 8 percent of Canadians said they knew of any AI rules or policies in their country, although 75 percent said regulation is needed.
- The numbers describe what people say they believe and have seen, not measured effects of AI on society.
What was surveyed, and by whom
The study, titled "Trust, attitudes and use of artificial intelligence: A global study 2025," was designed, run, and analyzed by a University of Melbourne team led by Nicole Gillespie and Steve Lockey. It sits within a research partnership between the university and KPMG Australia, and the report says KPMG funding supported it while the university team kept control of the research. The citable record carries the DOI 10.26188/28822919.
The survey was online. It covered 47 countries and jurisdictions, including both Canada and the United States, with 1,001 to 1,098 respondents in each. Each national sample was built to match official statistics on age, gender, and region within a 5 percent margin. The report also notes that online panels tend to overrepresent better-educated respondents in emerging economies, which matters less for a comparison between Canada and other advanced economies than for a comparison with, say, Nigeria or Egypt.
Canada beside the global average
The Canadian snapshot and the global report use the same questions, which allows a direct comparison. The table pairs the two for the indicators where both documents print a number. The Canadian column comes from the country snapshot and the global column from the main report.
| Indicator | Canada | All 47 countries |
|---|---|---|
| Willing to trust AI systems | 34% | 46% |
| Accept or approve of AI use | 50% | 72% |
| Received any AI training | 24% | 39% |
| Believe risks outweigh benefits | 46% | 32% |
| Say AI regulation is needed | 75% | 70% |
| Want laws against AI-generated misinformation | 88% | 87% |
The global report describes a split by income level: people in advanced economies were less trusting (39 percent) and less accepting (65 percent) than people in emerging economies (57 and 84 percent). Canada sits on the cautious side of that divide: its 50 percent acceptance is 15 points under the advanced-economy average and 22 points under the all-country figure.
What respondents said AI is doing for them and to them
On the benefit side, 60 percent of Canadians said they had personally experienced or observed benefits from AI, and the benefit listed first in the snapshot was reduced time on mundane or repetitive tasks, at 74 percent. On the risk side, 39 percent said they had experienced or observed negative outcomes, and the top item, at 60 percent, was a loss of human interaction and connection.
Two findings concern information rather than jobs or convenience. Seventy-seven percent said they are unsure whether online content can be trusted because it may be AI-generated, and 62 percent were concerned that elections could be manipulated by AI-generated content or bots. Respondents also wanted something done: 88 percent supported laws and action against AI-generated misinformation.
Regulation shows the sharpest distance between opinion and knowledge. Three in four Canadians said AI regulation is required, only 27 percent believed current safeguards are sufficient, and 8 percent said they were aware of AI regulations or policies in Canada. The study cannot say whether the 92 percent who were unaware are simply uninformed or are correct that few rules apply to them. That question belongs to legal analysis, and the sources for it are the statutes and guidance themselves.
Limits to keep in mind
A trust percentage depends on wording. "Willing to trust" and "accept or approve" are separate items, and they produced a 16-point gap in Canada. Sampling is another limit: online panels exclude people who are rarely online, and a sample of about 1,000 gives a comfortable margin for national figures but a much wider one for any subgroup.
The study also measures perception. A respondent who says AI has reduced human interaction is reporting an experience, and a different design would be needed to measure interaction time. Finally, the data are dated. Fieldwork ended in January 2025, before several product launches and policy changes of that year, so the figures should be read as one reading on a moving scale. The report itself compares 17 countries surveyed in both 2022 and 2024 and treats the changes as indicative trends, because the two samples were drawn independently.
Questions to ask before quoting a poll about AI
- Who ran the fieldwork, who paid for it, and who controlled the analysis?
- When was it collected, and has anything major changed since?
- Is the figure an opinion, a reported experience, or a measurement?
- What was the exact question, and what share answered "unsure" or "balanced"?
- How large is the subgroup being quoted?
FAQ
Does the survey show that AI is harming Canadian society?
No. It shows what respondents say they believe and have experienced. Perceived loss of human interaction is a reported experience, and the survey did not measure social contact directly.
Why do Canadians look more cautious than the global average?
The study's own explanation is a split between advanced and emerging economies, where trust and acceptance are lower in the former. It does not isolate Canada-specific causes, so any more detailed reason would be speculation.
Were Americans surveyed in the same study?
Yes, the United States is one of the 47 countries. This article uses Canadian figures only, because the printed U.S. figures could not be confirmed from the documents opened for it.
Is 46 percent a majority view that AI is risky?
No. In Canada 46 percent said risks outweigh benefits, which is more than the global 32 percent but less than half. The remaining respondents leaned toward benefits or called the two balanced.
Related reading
- Evidence on AI's impact by area
- Two social benefits backed by public records
- What federal rules ask of high-impact AI
- Why model versions and dates matter when comparing chatbots
More in AI Basics.
Last verified: October 2026
Sources
- Trust, attitudes and use of artificial intelligence: A global study 2025, Gillespie N., Lockey S., Ward T., Macdade A., Hassed G., University of Melbourne and KPMG, 2025
- Trust, attitudes and use of artificial intelligence: A global study 2025, Canada insights, University of Melbourne and KPMG, 2025