《墨爾本大學&畢馬威：2025全球人工智能信任、態度與應用研究報告（英文版）（118頁）.pdf》由會員分享，可在線閱讀，更多相關《墨爾本大學&畢馬威：2025全球人工智能信任、態度與應用研究報告（英文版）（118頁）.pdf（118頁珍藏版）》請在三個皮匠報告上搜索。

1、Trust,attitudes and use of artificial intelligenceA global study 2025University of Melbourne|KPMG Internationalunimelb.edu.au| CitationGillespie,N.,Lockey,S.,Ward,T.,Macdade,A.,&Hassed,G.(2025).Trust,attitudes and use of artificial intelligence:A global study 2025.The University of Melbourne and KPM

2、G.DOI 10.26188/28822919.Trust,attitudes and use of artificial intelligence:A global study is provided under a Creative Commons Attribution,Non-Commercial,Share Alike 4.0 International licence.You are free to use,share,reproduce and distribute the work under this licence for non-commercial purposes o

3、nly,as long as you give appropriate credit to the original author(s)and the source via the citation.If any changes are made to the material,information,graphics,etc,contained in this report,the changes must be clearly indicated.Under this licence,you may not use the material for any commercial purpo

4、ses.Any re-sharing of this material can only be done under the CC NC SA licence conditions.University of Melbourne Research Team Professor Nicole Gillespie,Dr Steve Lockey,Alexandria Macdade,Tabi Ward,and Gerard Hassed.Professor Nicole Gillespie and Dr Steve Lockey from the University of Melbourne l

5、ed the design,conduct,data collection,analysis,and reporting of this research.At various stages of the project,the research team sought feedback and input from a multidisciplinary advisory board,including academics and industry experts,while maintaining independence over the conduct and reporting of

6、 the research.AcknowledgmentsAdvisory group:James Mabbott,Jessica Wyndham,Nicola Stone,Sam Gloede,Dan Konigsburg,Sam Burns,Kathryn Wright,Melany Eli,Rita Fentener van Vlissingen,David Rowlands,Laurent Gobbi,Rene Vader,Adrian Clamp,Jane Lawrie,Jessica Seddon,Ed OBrien,Kristin Silva,and Richard Boele.

7、We are grateful for the insightful expert input and feedback provided at various stages of the research by Ali Akbari,Nick Davis,Shazia Sadiq,Ed Santow,Jeannie Paterson,Llewellyn Spink,Tapani-Rinta-Kahila,Alice Rickert,Lucy Kenyon-Jones,Morteza Namvar,Olya Ohrimenko,Saeed Akhlaghpour,Chris Ziguras,S

8、am Forsyth,Greg Dober,Giles Hirst,and Madhava Jay.We appreciate the data analysis support provided by Jake Morrill.Report production:Kathryn Wright,Melany Eli,Bethany Fracassi,Nancy Stewart,Yong Dithavong,Marty Scerri and Lachlan Hardisty.FundingThis research was supported by the Chair in Trust rese

9、arch partnership between the University of Melbourne and KPMG Australia,and funding from KPMG International,KPMG Australia,and the University of Melbourne.The research was conducted independently by the university research team.Trust,attitudes and use of AI:A global study 2025|2 2025 Copyright owned

10、 by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.2025 The University of Melbourne.ContentsList of figures 2Executive summary 4Introduction 11How the research was conducted 13Section 1:Public attitudes towards AI 18 To w

11、hat extent do people use and understand AI systems?19 To what extent do people trust and accept AI systems?27 How do people view and experience the benefits and risks of AI?37 What do people expect from the regulation and governance of AI?47 What are the key drivers of trust and acceptance of AI sys

12、tems?59 How do demographic factors influence trust,attitudes and use of AI?62Section 2:Employee attitudes towards AI at work 66 How is AI being used by employees at work?67 What are the impacts of AI use at work?77 How do demographic factors influence use and perceptions of AI at work?85Section 3:St

13、udent attitudes towards AI in education 89 How is AI being used by students?90 What are the impacts of AI use in education?93Conclusion and implications 96Appendix 1:Methodological and statistical notes 104Appendix 2:Sample demographics 107Appendix 3:Key indicators for each country 109Appendix 4:Cha

14、nges in key indicators over time for 17 countries 110Trust,attitudes and use of AI:A global study 2025|1 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.List of figu

15、resFigure 1:Frequency of intentional use of AI tools for personal,work,or study purposes 20Figure 2:Use of AI systems on a regular or semi-regular basis across countries 21Figure 3:AI-related training or education 22Figure 4:Self-reported AI knowledge 22Figure 5:Self-reported AI efficacy 22Figure 6:

16、AI training and education,knowledge and AI efficacy across economic groups 23Figure 7:AI knowledge,efficacy,and training across countries 24Figure 8:Use of common technologies and awareness that they involve AI 25Figure 9:Perceptions of the trustworthiness of AI systems 28Figure 10:Trust and accepta

17、nce of AI systems 29Figure 11:Trust in AI applications across countries 30Figure 12:Trust and acceptance of AI systems across economic groups 31Figure 13:Trust and acceptance of AI systems across countries 32Figure 14:Emotions associated with AI 33Figure 15:Emotions toward AI across countries 34Figu

18、re 16:Trust of AI systems and worry about AI in 2022 and 2024 35Figure 17:Expected and experienced benefits of AI use 38Figure 18:Expected benefits of AI across countries 39Figure 19:Experienced benefits of AI across countries 40Figure 20:Perceived risks and experienced negative outcomes from AI use

19、 41Figure 21:Concerns about the risks of AI across countries 43Figure 22:Experienced negative outcomes from AI use across countries 44Figure 23:Perceptions across countries that AI benefits outweigh risks 45Figure 24:Need for AI regulation across countries 49Figure 25:Perceived adequacy of current r

20、egulation and laws to make AI use safe 50Figure 26:Expectations of who should regulate AI 51Figure 27:Expectations of who should regulate AI across countries 52Figure 28:Impacts and management of AI generated misinformation 53Figure 29:AI assurance mechanisms 54Figure 30:Confidence in entities to de

21、velop and use AI 56Figure 31:Confidence in entities to develop and use AI across countries 57Trust,attitudes and use of AI:A global study 2025|2 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to

22、clients.All rights reserved.List of figures contdFigure 32:A model of the key drivers of trust and acceptance of AI use in society 60Figure 33:Trust and acceptance of AI systems by age,income,education,and AI training 64Figure 34:Use of AI and AI training by age,income,and education 64Figure 35:AI k

23、nowledge and AI efficacy by age,income,and education 65Figure 36:Organizational use of AI(employee reported)67Figure 37:Frequency of intentional use of AI at work 68Figure 38:Organizational and employee AI adoption have increased over time 69Figure 39:Types of AI tools intentionally used at work 70F

24、igure 40:Access to AI tools used at work 71Figure 41:Organizational policy or guidance on generative AI at work(employee reported)71Figure 42:Frequency of intentional use of AI at work 72Figure 43:Intentional use of AI at work and trust of AI at work 73Figure 44:Inappropriate and complacent use of A

25、I at work 76Figure 45:Critical engagement with AI at work 76Figure 46:Impacts of AI use in the workplace as reported by employees 78Figure 47:Employee reliance on AI at work 79Figure 48:Preference for humanAI involvement in managerial decision-making 79Figure 49:Perceived organizational support for

26、AI and responsible AI use 81Figure 50:Organizational support for AI and responsible use across countries 82Figure 51:Perceived impact of AI on jobs 83Figure 52:Demographic differences in trust and use of AI at work 87Figure 53:Demographic differences in complacent use and positive impacts of AI 87Fi

27、gure 54:Industry differences in use of AI and organizational support for AI 88Figure 55:Frequency of student use of AI compared to employee use of AI for work 90Figure 56:Types of AI tools intentionally used for study,compared to employees 91Figure 57:Inappropriate and complacent use of AI in educat

28、ion 92Figure 58:Impacts of AI use in education as reported by students 94Figure 59:Education provider support for responsible AI use as reported by students 95Figure 60:Education providers guidance on generative AI use for students 95Trust,attitudes and use of AI:A global study 2025|3 2025 The Unive

29、rsity of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.Executive summaryThe release of ChatGPT in late 2022 brought the transformative power of AI firmly into the public consciousness an

30、d everyday experience.While exponential investment in AI predated its release,individual and organizational use of AI has increased dramatically and rapidly since 2022.1 For example,OpenAIs suite of generative AI tools obtained over 100 million users in only two months.2 AI is now firmly part of eve

31、ryday life and work for many people and is widely embraced across all sectors of the global economy,including finance,education,transport,manufacturing,agriculture,healthcare,retail,and media.3 The benefits and promise of AI for society and business are undeniable.AI systems are being used to make c

32、ancer detection faster and more accurate,enhance the efficiency of renewable energies,and drive productivity and innovation in the workplace,among other impactful use cases.4 However,as AIs capabilities and reach become more apparent,so too has awareness of the risks and challenges,raising questions

33、 about the trustworthiness,regulation,and governance of AI systems.The publics trust in AI technologies and its responsible and ethical use is central to sustained acceptance and adoption and in realizing the full societal and economic benefits of these technologies.Given the rapid advancement and w

34、idespread adoption of AI technologiesand their transformative effects on society,work,education,and the economybringing the public voice into the conversation has never been more critical.This research aims to provide an evidence-based understanding of peoples trust,use and attitudes toward AI,their

35、 views on the impacts of AI,and expectations of its governance and regulation.The insights are important to inform public policy and industry practice and a human-centered approach to stewarding AI into work and society.They can help policymakers,organizational leaders,and those involved in developi

36、ng,deploying,and governing AI systems to understand and align with evolving public expectations,and deepen understanding of the opportunities and challenges of AI integration.The report provides timely,global research insights on a range of questions,including the extent to which people trust,use,an

37、d understand AI systems;how they perceive and experience the benefits,risks and impacts of AI use in society,at work and in education;expectations for the management,governance and regulation of AI by organizations and governments;how employees and students are using AI for work and study;and percei

38、ved support for the responsible use of AI.It draws out commonalities and differences in these key dimensions across countries and sub-groups of the population,and sheds light on how trust and attitudes toward AI have changed over the past two years since the widespread uptake of generative AI.Next,w

39、e summarize the key research insights.Now in its fourth iteration,the research captures the views of more than 48,000 people from 47 countries,representing all global geographic regions.It offers the most comprehensive examination to date of public trust and attitudes toward AI.In addition,it takes

40、a deep dive into how employees and students use AI in work and education and their experience of the impacts of AI in these specific settings.Trust,attitudes and use of AI:A global study 2025|4 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KP

41、MG International entities provide no services to clients.All rights reserved.A snapshot of key findingsTrust and acceptance of AITrust in AI systems remains a significant challenge:over half(54%)are wary about trusting AI.People are more skeptical of the safety,security and societal impact of AI and

42、 more trusting of its technical ability.While most people feel both optimistic and worried about AI,72%accept its use.People in advanced economies are less trusting(39%vs.57%)and accepting(65%vs.84%)compared to emerging economies.AI use and understandingTwo in three(66%)intentionally use AI on a reg

43、ular basis and three in five say they can use AI effectively.However,most(61%)have no AI training and half report limited knowledge.People in emerging economies report higher regular use(80%vs 58%),training(50%vs 32%),knowledge(64%vs 46%)and efficacy(74%vs 51%)than those in advanced economies.People

44、 that are younger,university-educated,higher-income earners and AI-trained report more trust,use and AI literacy.AI benefits and risks People report experiencing both benefits and negative outcomes from AI use.While many report improved efficiency,accessibility,decision-making and innovation,concern

45、s about cybersecurity,privacy and IP,misinformation,loss of human connection,job loss and deskilling are widespread.The publics ambivalence towards AI is evident,with divided opinion on whether the benefits outweigh the risks in advanced economies.AI regulation and governance There is a strong publi

46、c mandate for AI regulation,with 70%believing regulation is necessary.However,only 43%believe current laws are adequate.People expect international laws(76%),national government regulation(69%),and co-regulation with industry(71%).87%also want laws and fact-checking to combat AI-generated misinforma

47、tion.AI adoption in the workplace Three in five(58%)employees intentionally use AI at work on a regular basis,with a third using it weekly.Generative AI tools are most commonly used with many employees opting for free,publicly available tools rather than employer-provided options.Emerging economies

48、are leading in employee adoption with 72%using AI regularly compared to 49%in advanced economies.Impacts of AI at workOver half of employees report performance benefits from AI.However,employees also report mixed impacts on workload,human interaction and compliance and two in five believe AI will re

49、place jobs in their area.Many employees report inappropriate,complacent and non-transparent use of AI in their work,contravening policies and resulting in errors and dependency.Governance and training to support responsible AI use appears to be lagging adoption.Student engagement with AI Four in fiv

50、e students(83%,predominately tertiary)regularly use AI in their studies,reporting benefits such as efficiency,personalization of learning,and reduced workload and stress.However,inappropriate,complacent and non-transparent use of AI by students is widespread,raising concerns about over-reliance and

51、diminished critical thinking,collaboration,and equity of assessment.Only half report their education provider has policies,resources or training to support responsible AI use.Trust,attitudes and use of AI:A global study 2025|5 2025 The University of Melbourne.2025 Copyright owned by one or more of t

52、he KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.The age of working with AI is here and is delivering performance benefits,but also mixed impactsAcross countries,almost three in five employees intentionally5 use AI at work on a regular bas

53、is,with almost a third using it weekly or more.General-purpose generative AI tools are by far the most widely used,with most employees using free,public tools like ChatGPT rather than tools provided by their employer.Three in four report that their organization uses AI,with almost half stating AI is

54、 used in a broad range of tasks and functions.Emerging economies6 are leading workplace adoption of AI,with employees in these economies more likely to use AI regularly(72%vs 49%)than those in advanced economies.The use of AI at work is clearly delivering a range of positive performance benefits.Mos

55、t employees report increased work efficiency,access to accurate information,innovation,higher quality of work and decisions,and better use and development of skills and abilities.Almost half report that AI use has increased revenue-generating activity.However,employees also report mixed impacts on w

56、orkload,stress,human collaboration,compliance,and surveillance at work.For example,half say they use AI rather than collaborating with peers or supervisors to get work done,and one in five say AI use has reduced communication,interaction and collaboration,raising the question of how human connectivi

57、ty will be retained in AI-augmented workplaces.These insights underscore the importance of understanding and managing the impacts of AI at work,ensuring appropriate work design,and building employee capabilities in effective human-AI collaboration.The responsible use and governance of AI is not keep

58、ing pace with adoption:many employees are using AI in complacent and inappropriate ways which increase riskWhile the rapid adoption of AI is delivering benefits,many employees are using AI in complacent and inappropriate ways,increasing risks for organizations and individuals and raising quality iss

59、ues.For example,almost half admit to using AI in ways that contravene organizational policies and uploading sensitive company information,such as financial,sales,or customer information,to public AI tools.Three in five report they have seen or heard of other employees using AI tools in inappropriate

60、 ways.Two in three report relying on AI output without evaluating the information it provides,and over half say they have made mistakes in their work due to AI.What makes these risks even more challenging to manage is that over half of employees avoid revealing when they use AI to complete their wor

61、k and present AI-generated content as their own.These findings highlight a lack of transparency and accountability in the way AI,particularly generative AI tools,are being used by employees at work.This complacent use may be fueled by inadequate training,guidance,and governance of responsible AI use

62、 at work:within organizations that use AI,only one in two employees in advanced economies report that their organization offers training in responsible AI,has policies and practices on responsible AI use,or a strategy and culture that supports AI.Despite the high use of generative AI tools,only two

63、in five say there is a policy guiding its use.Complacent use may also be exacerbated by a sense of pressure to use AI,with half of employees feeling they will be left behind if they dont.From a governance perspective,these findings highlight a critical gap and urgent need for organizations to proact

64、ively invest in responsible AI training and the AI literacy of employees to promote critical engagement with AI tools.They also underscore the need to put in place mechanisms to effectively guide and govern how employees use AI tools in their everyday work,to promote greater accountability,transpare

65、ncy,and employee engagement.Trust,attitudes and use of AI:A global study 2025|6 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.Most students use AI and report benef

66、its,but inappropriate use and over-reliance is widespread and challenging critical skill development The findings for students(predominately tertiary students)provide insight into how AI is being used by the next generation of the workforce and affecting education and training.Results mirror those f

67、or employees but are more pronounced.Four in five(83%)students regularly use AI in their studies,with half using it weekly or daily.The large majority use free,publicly available generative AI tools.Most students are deriving significant benefits from AI use in education,such as increased efficiency

68、,access to information,quality of work,idea generation and personalization of learning,and reduced workload and stress.However,AIs influence on social dynamics,critical thinking,and assessment is mixed.For example,a quarter to a third of students report reduced critical thinking and less communicati

69、on,interaction,and collaboration with instructors and peers.A similar number perceive less trust of students by instructors and peers,and reduced fairness and equity of assessment due to AI.The complacent use of AI by students is widespread.Most students have used AI inappropriately,contravening rul

70、es and guidelines and over-relying on AI.Two-thirds have not been transparent in their AI use,presenting AI-generated content as their own and hiding their use of AI tools.Only half regularly engage critically with AI tools and their output.The level of student dependence on AI is concerning:over th

71、ree-quarters have felt they could not complete their work without the help of AI and rely on it to do tasks rather than learning how themselves.Four in five say they put less effort into their studies and assessment knowing they can rely on AI.A lack of institutional support for responsible AI use m

72、ay be contributing to this problem:only half of students report their education provider has policies to guide responsible use of AI in learning and assessment,or training and resources to support AI understanding and responsible use.These findings may have longer-term implications for the effective

73、 development of essential skillssuch as critical thinking,communication and collaboration,with implications for organizations as these students enter the workforce.Trust in AI cannot be taken for granted:many people are wary about trusting AI systems,particularly in advanced economiesDespite high ra

74、tes of individual adoption,trust remains a critical challenge.Over half(54%)of people are wary about trusting AI systems.Underlying this average are differences between economic groups:three in five people in advanced economies are unwilling or unsure about trusting AI systems.In contrast,in emergin

75、g economies,three in five people trust AI systems.We find similar levels for employee trust in the use of AI at work,and student trust of AI for educational purposes.People are more skeptical about the safety,security,and ethical use of AI systems and more trusting of the technical ability of AI to

76、provide helpful output and services.This helps explain individual use of AI to gain performance benefits,despite trust concerns around its broader impact on society and people.While the majority accept the use of AI systems,most people report low or moderate acceptance and approval levels.Peoples am

77、bivalence toward AI is also reflected in their emotions:the majority report optimism and excitement,coupled with worry.People have high confidence in universities,research,and healthcare institutions to use and develop AI in the best interests of the public,and generally less confidence in governmen

78、t to do so.People in advanced economies have lower confidence in industry and big technology companies to develop and use AI in the public interest,whereas confidence in these entities is high in emerging economies.Organizations can build stakeholders trust in their use of AI by investing in respons

79、ible AI governance mechanisms that signal trustworthy use:four in five people report they would be more willing to trust an AI system when assurance mechanisms are in place,such as monitoring system reliability,human oversight and accountability,responsible AI policies and training,adhering to inter

80、national AI standards,and independent third-party AI assurance systems.Trust,attitudes and use of AI:A global study 2025|7 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights res

81、erved.People are experiencing a range of benefits and negative outcomes from the use of AI in society Peoples ambivalence toward AI stems from the mixed benefits,risks and negative impacts that are being felt from AI use in society:42 percent believe the benefits outweigh the risks,32 percent believ

82、e the risks outweigh the benefits,and 26 percent believe the benefits and risks are balanced.Three in four report experiencing a broad range of benefits,including improved efficiency and effectiveness,enhanced accessibility to information and services,greater precision and personalization,improved d

83、ecision-making and outcomes,greater innovation and creativity,reduced costs and better use of resources.These outcomes benefit individuals,while also bringing performance-oriented benefits to organizations and society more broadly.However,peoples experience of these benefits is coupled with clear co

84、ncerns about the risks and negative impacts of AI on society.Four in five people are concerned aboutand two in five have personally experienced or observednegative outcomes from AI.These include the loss of human interaction and connection,cybersecurity risks,loss of privacy or intellectual property

85、,misinformation and manipulation,harmful or inaccurate outcomes,deskilling and dependency,job loss,and disadvantage from unequal access to AI.Comparatively fewer people are concerned about AI bias resulting in unfair treatment and the environmental impact of AI,however even these outcomes are report

86、ed by a third of people surveyed.Respondents across countries share similar views and experiences regarding AI risks and negative outcomes,highlighting these as areas of universal concern.These negative outcomes are not just perceived risks but harms that are being experienced or observed by a signi

87、ficant proportion of people across the 47 countries surveyed.These findings reinforce the need for international cooperation and coordinated action to prevent and mitigate AI risks and negative impacts.The challenge is doing this in a balanced way that does not undermine progress or hinder the innov

88、ation required to realize the many societal benefits of AI.The public expect AI regulation at both the national and international level.Yet the current regulatory landscape is falling short of public expectations.There is a strong public mandate for AI regulation to mitigate the societal risks and n

89、egative impacts of AI:Seventy percent of people believe AI regulation is required,including the majority in almost all countries surveyed.This broad public consensus on the need for regulation supports national and international efforts in many jurisdictions to develop and implement regulatory and g

90、overnance frameworks to support the safe and responsible use of AI.However,the current regulatory landscape is falling short of public expectations:only two in five believe that the existing laws and regulation governing AI systems in their country are adequate.Most people are unaware of laws,legisl

91、ation or government policy that apply to AI.These findings reflect that most countries and jurisdictions are still in an early stage of designing or implementing regulatory approaches.While some countries have adaptive legislation that may apply to AI(e.g.consumer or privacy laws),such laws are abse

92、nt or weakly enforced in some jurisdictions.This suggests the need to clarify,develop or strengthen such legislation where it is lacking and to educate and raise public awareness of applicable laws.The importance of effective,fit-for-purpose regulationand awareness of such regulationis underscored b

93、y our finding that the perceived adequacy of AI regulation is a key predictor of trust and acceptance of AI systems.The majority of people expect a multipronged national and international regulatory approach to AI,with international laws and regulation the most endorsed form of regulation and suppor

94、ted by a clear majority in all countries.National government regulation or a co-regulatory approach between government and industry is preferred in most countries over self-regulation by industry or an independent AI regulator.This highlights the publics expectation that government takes a central r

95、ole in ensuring effective governance and regulation of AI,as well as the expectation that industry will work with regulatory bodies and proactively align their governance approach with the evolving regulatory landscape.Trust,attitudes and use of AI:A global study 2025|8 2025 The University of Melbou

96、rne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.There is also a clear mandate for stronger regulation of AI-generated misinformation:87 percent of respondents want laws to combat AI-generated mi

97、sinformation and expect social media and media companies to implement stronger fact-checking processes and methods that enable people to detect AI-generated content.Our findings indicate that AI generated misinformation is a key concern globally and is undermining trust in online content and raising

98、 concerns about the integrity of elections.AI literacy is lagging AI adoption yet is critical for responsible and effective useAlthough AI tools are being widely used by the public,employees and students,AI literacy remains limited;about half of respondents say they dont feel they understand AI nor

99、when or how it is used.Half of respondents are unaware that AI underpins common applications such as social media,despite 90 percent saying they use such platforms.This knowledge gap reflects that only two in five people report any AI-related training or education.Despite low rates of knowledge and

100、training,three in five say they can use AI effectively.This likely reflects the easily accessible interfaces of many AI systems(e.g.using natural language)and low barriers to use.While this accessibility has benefits,it also risks fostering complacency and overreliance if not accompanied by meaningf

101、ul levels of understanding and literacy.AI literacy is higher in emerging economies,where three-quarters believe they can use AI effectively,compared to half in advanced economies,and half report AI training or education compared to a third in advanced economies.AI literacy consistently emerges in o

102、ur findings as a cross-cutting enabler:it is associated with greater use,trust,acceptance,and critical engagement,and more realized benefits from AI use including more performance benefits in the workplace.The pattern of findings underscores that AI literacy and training in responsible use is not on

103、ly a personal skillset,but can also be a strategic capability for organizations and societies alike,enabling people to recognize and seize the capabilities of AI while recognizing their limitations and guarding against harm.Investing in AI literacy is a critical component of ensuring AI is used safe

104、ly,ethically,and to its full potential.There are notable differences between countries with advanced and emerging economies:People in emerging economies report greater trust,acceptance and adoption of AI,higher levels of AI literacy,and more realized benefits from AIOne of the most striking insights

105、 from the survey is the stark contrast in use,trust,and attitudes toward AI between people in advanced and emerging economies.People in emerging economies report higher adoption and use of AI both at work and for personal purposes,are more trusting and accepting of AI,and feel more positive about it

106、s use.They report higher levels of AI training and literacy,are more likely to expect and realize the benefits of AI,and view AI benefits as outweighing the risks.They are also more confident in the development and use of AI by commercial organizations and big technology companies and more likely to

107、 view current AI regulation and safeguards as adequate,compared to people in advanced economies.These differences hold even when controlling for the effects of age and education.These findings suggest that many countries with emerging economies are leading the way in terms of AI adoption.7 In partic

108、ular,six countries with emerging economies strongly and consistently show this patternIndia,China,Nigeria,the UAE,Saudi Arabia and Egypt.Of the advanced economies,Norway,Israel,Singapore,Switzerland and Latvia have comparatively high levels of AI adoption,trust,acceptance,and positive attitudes towa

109、rd AI.Trust,attitudes and use of AI:A global study 2025|9 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.An implication is that these countries may be uniquely posi

110、tioned to rapidly accelerate innovation and technological advantage through AI.This has implications for global competitive dynamics and may create shifts in the economic landscape across countries in the future as AI becomes a more prominent driver of productivity and economic activity.Pathways to

111、support the trusted and responsible adoption of AI Our modeling supports four distinct yet complementary pathways to trusted and sustained AI adoption:a knowledge pathway reflecting the importance of supporting peoples AI literacy and efficacy through AI training and education;a motivational pathway

112、 reflecting the importance of deploying AI in a human-centric way that delivers benefits to people;the uncertainty reduction pathway reflecting the need to address concerns about the risks associated with AI,and an institutional pathway reflecting the adequacy of current safeguards,regulation and la

113、ws to promote safe AI use,and confidence in entities to develop and use AI in the public interest.Of these drivers,the institutional pathway had the strongest influence on trust,followed by the motivational pathway.This model also holds at the organizational level where the institutional pathway ref

114、lects appropriate levels of organizational governance,strategy,and training to support AI and its responsible use.AI adoption has increased markedly since 2022,but trust in AI has declined and worry has increasedOur research program provided the unique opportunity to compare data from the current su

115、rvey with our previous survey data collected from 17 countries in late 2022,just prior to the release of ChatGPT.This comparison revealed a trend of less positive attitudes toward AI,as adoption has increased.As expected,adoption of AI in the workplace increased dramatically in all 17 countries:empl

116、oyee reported organizational use of AI increased from 34 percent to 71 percent,and employees use of AI at work increased from 54 percent to 67 percent.The largest increases occurred in Australia,Canada,the USA,and the UK.However,this increased adoption is coupled with a trend toward people feeling m

117、ore concerned about and less trusting of AI.Peoples perceptions of the trustworthiness of AI systems and their willingness to rely on AI declined in most countries,as did employee trust of AI at work in some countries.This decline in trust likely reflects that increased use and exposure,particularly

118、 to general-purpose generative AI tools,has increased awareness of both the capabilities and benefits of these tools,and also their limitations and potential negative impacts(e.g.hallucinations),prompting more considered trust and reliance.More people report feeling worried about AI and concerned ab

119、out the risks,and fewer view the benefits of AI as outweighing the risks.For example,in Brazil half of people reported feeling worried about AI in 2022 compared to 75%in 2024,and the view that the benefits of AI outweigh the risks fell from 71%to 44%.Excitement also dampened over this time in severa

120、l countries.With this increase in concern,the importance of organizational assurance mechanisms as a basis for trust increased in all countries,suggesting a greater need for reassurance that AI is being used in a trustworthy and responsible way.Attitudes toward the regulation of AI remained stable a

121、nd there was no overall change to the perceived adequacy of regulation and laws.Despite the rapid uptake of AI,we found no discernible change in the publics self-reported understanding of AI,or their objective awareness of AI use in common applications.This pattern of findings suggests that the hype

122、 of AI may be giving way to a more realistic and measured assessment of AIs capabilities and limitations,benefits and risks,and heightened need for reassurance around the trustworthy deployment of AI and proactive mitigation of AI risks.Collectively,the survey insights provide evidence-based pathway

123、s for strengthening the responsible use of AI systems and the trusted adoption of AI in society and work.These insights are relevant for informing responsible AI strategy,practice and policy within business,government,and education at a national level,as well as informing AI guidelines,policy and re

124、gulation at the international and pan-governmental level.Trust,attitudes and use of AI:A global study 2025|10 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.This is

125、 the fourth survey in our program of research examining public trust and attitudes toward AI.Our current report examines the perspectives of over 48,000 people from 47 countries covering all global geographic regions,using nationally representative sampling of the adult population based on age,gende

126、r,and regional distribution.Taking a global perspective is crucial,given that AI systems are not bound by physical borders and are rapidly being deployed and used across the world.Our program of research provided the unique opportunity to benchmark and compare the findings in this report to our prev

127、ious survey data collected from 17 countries in late 2022,just prior to the release of ChatGPT.We examine changes in public trust and attitudes over time in these 17 countries and highlight changes where relevant throughout the report(see How we conducted the research for more details).IntroductionT

128、he motivation for this research is to provide an evidence-based understanding of public trust,attitudes,and experiences of AI,and expectations of its governance and regulation,as a resource to inform public policy and industry and government practice.Given the rapid advancement,widespread deployment

129、 and transformative impact of AI technologies,it is important to regularly examine public trust,attitudes,and expectations of AI.Equally important is documenting how people use AI technologies and experience the impacts of AI in their lives,work,and studies,and the implications this may have for org

130、anizations,education providers,and society at large.To date,there has been limited empirical insight addressing these critical issues,underscoring the relevance of this research in promoting human-centered AI that meets evolving societal needs and expectations.The Trust in AI Research ProgramThis st

131、udy is the fourth in a research program examining public trust in AI.Each study has been designed to uphold academic rigor and independence,whilst leveraging the deep multidisciplinary expertise and insight from KPMG.The first focused on Australians trust in AI in 2020,the second expanded the resear

132、ch scope to study trust in five countries in 2021,and the third surveyed people in 17 countries in 2022.Trust,attitudes and use of AI:A global study 2025|11 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no

133、services to clients.All rights reserved.Our research insights are structured in three sections.The first focuses on AI use broadly in society examining the publics use,understanding,trust,attitudes and experience of AI systems and their impact on society.These insights are based on all respondents a

134、nswering survey questions asked about AI systems in general,as well as AI use in the context of three common applications which are likely to be used by or impact many people:generative AI systems,AI in healthcare,and AI in Human Resource applications.In the second section,we delve deeper into under

135、standing how employees use and experience AI impacts in the workplace.In the third section,we examine student use of AI and their perceptions of how AI impacts education.Together,these sections provide evidence-based insights on the following questions:To what extent do people use and understand AI

136、systems?To what extent do people trust and accept AI systems?How do people view and experience the benefits and risks of AI?What do people expect from the regulation and governance of AI?What are the key drivers of AI trust and acceptance in society?How is AI being used at work and with what impacts

137、?How is AI being used by students and with what impacts?The final section draws out the key conclusions and implications from these insights for industry,government,and the education sector.We next outline the research methodology.Trust,attitudes and use of AI:A global study 2025|12 2025 The Univers

138、ity of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.How the research was conducted48,340 people completed the survey across 47 countries and jurisdictions,covering all global geographic

139、al regions8:1.North America(Canada,United States of America USA)2.Latin America and Caribbean(Argentina,Brazil,Chile,Colombia,Costa Rica,Mexico)3.Northern and Western Europe(Austria,Belgium,Denmark,Estonia,Finland,France,Germany,Ireland,Latvia,Lithuania,Netherlands,Norway,Sweden,Switzerland,United K

140、ingdom UK)4.Southern Europe(Greece,Italy,Portugal,Slovenia,Spain)5.Eastern Europe(Czech Republic,Hungary,Poland,Romania,Slovakia)6.Africa(Egypt,Nigeria,South Africa)7.Western Asia(Israel,Saudi Arabia,Trkiye,United Arab Emirates UAE)8.Eastern,Southern and Central Asia(China,9,India,Japan,Republic of

141、Korea,Singapore)9.Oceania(Australia,New Zealand)How the data was collected Data was collected in each country between November 2024 and mid-January 2025 using an online survey.Countries were selected based on three criteria:1)representation across global regions;2)leadership in AI activity and readi

142、ness,10 and 3)diversity on the Responsible AI Index.11 The sample size in each country ranged from 1,001 to 1,098 respondents.Analysis of the data revealed a distinct pattern of findings across countries with emerging and advanced economies.We adopted the International Monetary Funds(IMF)classificat

143、ion of advanced and emerging economies.The emerging economies surveyed are Argentina,Brazil,Chile,China,12 Colombia,Costa Rica,Egypt,Hungary,India,Mexico,Nigeria,Poland,Romania,Saudi Arabia,South Africa,Trkiye,and UAE.Surveys were conducted in the native language(s)of each country with the option to

144、 complete in English,if preferred.To ensure question equivalence across countries,surveys were professionally translated and back translated from English to each respective language,using separate translators.See Appendix 1 for further method details.Trust,attitudes and use of AI:A global study 2025

145、|13 2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.2025 The University of Melbourne.ArgentinaAustraliaAustriaBelgiumBrazilCanadaChileCosta RicaChinaCzech Rep.ColombiaDenmarkEgyptEstoniaFinlandFranc

146、eGermanyGreeceHungaryItalyIndiaIrelandIsraelJapanLatviaLithuaniaMexicoNetherlandsNew ZealandNigeriaNorwayRepublic of KoreaPolandRomaniaPortugalSaudi ArabiaSingaporeSlovakiaSloveniaSouth AfricaSpainSwedenSwitzerlandUnited KingdomTrkiyeUnited States of AmericaUnited Arab EmiratesThe 47 countries surve

147、yedTrust,attitudes and use of AI:A global study 2025|14 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.Who completed the survey?Representative research panels were

148、used to ensure the people who completed the survey are representative of the population.13 This approach is common in survey research.Samples were nationally representative of the adult population on gender,age and regional distribution matched against official national statistics.In select countrie

149、s,full representation on these criteria was not obtainable(see Appendix 2 for further details on country sampling).Across the total sample,the gender balance was 51 percent women,49 percent men and 1 percent other gender identities.The mean age was 46 years and ranged between 18 and 95 years.Half th

150、e sample(51%)had a university education and 20 percent a vocational or trade qualification.The sample represented the full range of income levels,with the majority(72%)reporting middle incomes(see Appendix 1 for details of the income measure).14Sixty-seven percent of respondents were currently worki

151、ng full-time or part-time.These respondents represented the diversity of industries and occupational groups listed by the OECD and International Labor Organization15 and included employees of small,medium,and large organizations,business owners,and people who were self-employed(e.g.sole traders and

152、freelancers).Five percent of respondents were students,with the majority tertiary students enrolled in university education(65%)or a vocational,trade or technical program(16%),and the remainder in secondary education(18%).Further details of the sample representativeness,including the demographic pro

153、file for each country sample,are shown in Appendix 2.Gender51%Women49%MenOther genders1%Age Group12%18-2438%25-4432%45-6418%65-95Education2%Primary4%Some secondary23%Secondary20%Vocation/trade37%Undergraduate14%PostgraduateIncome Group15%72%13%HighMiddleLowWork Status52%28%Not working15%Working part

154、 timeWorking full timeEmployment TypeOrganization Size77%Employed by an organization7%Business owner with employees16%Self-employed26%Small(2-49 employees)32%Medium(50-249 employees)42%Large(250+employees)Current Education Program18%Secondary education16%Vocation or trade54%Bachelors or equivalent11

155、%Postgraduate1%Other5%Students(n=2,499)Occupation32%Professional&skilled22%Manager21%Administrative14%Manual10%Services&Sales1%Other67%Employees(n=32,352)Trust,attitudes and use of AI:A global study 2025|15 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG Internationa

156、l entities.KPMG International entities provide no services to clients.All rights reserved.How we asked about AIAfter asking a series of questions about respondents understanding of AI,the following description of AI,adapted from the OECD definition,16 was provided:Artificial Intelligence(AI)refers t

157、o machine-based systems that infer from the input they receive and objectives provided,how to generate outputs such as predictions,content,recommendations,or decisions.Different AI systems vary in their levels of autonomy and adaptiveness.As attitudes toward AI systems may depend on their purpose an

158、d use,survey questions that asked about the use of AI systems in society referred to one of four AI use cases(randomly allocated,see below):Generative AI(used to create output and content in response to user prompts);Healthcare AI(used to inform decisions about how to diagnose and treat patients);Hu

159、man Resources AI(used to inform decisions about hiring and promotion);and AI systems in general.These use cases were selected to represent AI applications that are widely and increasingly used and can impact many people,and were developed based on expert input.Respondents were provided with a descri

160、ption of the AI use case allocated to them,before answering questions related to AI systems.Generative AIA form of AI used to create content such as text,images,audio,and video based on user prompts.It works by processing these prompts and generating new content based on patterns and structures it h

161、as learned from extensive amounts of data.People use generative AI for a wide range of applications,such as writing,programming,personalized education,administrative support,product design and development,forecasting,and creating art and music.Human Resources AIAn AI system used to help select the m

162、ost suitable applicants for a job,identify employees who are most likely to perform well in a job,and predict who is most likely to quit.It works by collecting and comparing worker characteristics,employee data,and performance over time,and analyzing which qualities are related to better job perform

163、ance and job retention.Managers use Human Resources AI to inform decisions about hiring and promotion.Healthcare AIAn AI system used to improve the diagnosis of disease(e.g.cancer),inform the best treatment options,and predict health outcomes based on patient data.It works by comparing a patients he

164、alth data(e.g.symptoms,test and scan results,medical history,family history,age,weight and gender,etc.)to large datasets based on many patients.Doctors use Healthcare AI to inform decisions about patient diagnosis and treatment.Trust,attitudes and use of AI:A global study 2025|16 2025 The University

165、 of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.How the data was analyzedStatistical analyses were conducted to examine differences between countries and economic groups(e.g.countries

166、with advanced and emerging economies,as classified by the IMF),and demographic factors(e.g.gender,age,education,income,occupation).Relevant differences are reported when statistically significant and meaningful.Correlational analyses and statistical models indicate associations between concepts and

167、do not infer causality.Further details of the statistical procedures are discussed in Appendix 1.An overview of key indicators for each country sample are shown in Appendix 3.How changes in trust,use and attitudes over time were assessed To understand how trust,use,and attitudes toward AI have shift

168、ed over time,a selection of questions was asked in the same way in the 2022 and 2024 surveys.The 2022 survey included 17 countries:Australia,Brazil,Canada,China,Estonia,Finland,France,Germany,India,Israel,Japan,Netherlands,Singapore,South Africa,Korea,the UK,and the USA.17While the samples collected

169、 in 2022 and 2024 are based on the same methodology and sample representativeness,they are independent of each other.As such,our analyses examine general trends rather than a longitudinal analysis of the same respondents over time.Relevant insights on these changes are highlighted in call-out boxes

170、throughout the report(for an overview,see Appendix 4).Trust,attitudes and use of AI:A global study 2025|17 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.Public att

171、itudes towards AISECTION ONEIn this first section,we examine the publics adoption and understanding of AI and their trust,acceptance,and emotions towards the use of AI systems in society.We explore peoples expectations and experience of positive and negative impacts from AI systems,how they view the

172、 benefits relative to the risks,and expectations of AI regulation and governance.We test a model identifying key predictors of AI trust and acceptance and explore how people from various demographic groups differ in their attitudes toward and use of AI.Trust,attitudes and use of AI:A global study 20

173、25|18 2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.2025 The University of Melbourne.To what extent do people use and understand AI systems?To contextualize the findings and provide an indicator o

174、f overall public adoption of AI and AI literacy levels,we first examine peoples use and understanding of AI systems and how this varies across countries.To identify levels of AI literacy,survey participants self-reported their level of AI knowledge and efficacy together with AI-related education and

175、 training.They were also asked about their objective understanding of AI use in common technologies and interest in learning more about AI.In subsequent sections of the report,employees and students use of AI at work and for educational purposes are examined in more detail,together with organization

176、al support for AI literacy.Public adoption of AI is high:Two in three people report intentional regular use of AI tools for either personal,work,or study purposesPeople were asked to report how often they intentionally use AI tools,clarifying that this use is different from the passive use of AI(e.g

177、.when AI operates behind the scenes in tools such as email filters and search engines).Two thirds of people(66%)report intentionally using AI on a regular basis for personal,work,or study reasons.As shown in Figure 1,two in five(38%)people report using AI on a weekly or daily basis,whereas just over

178、 a quarter(28%)use AI semi-regularly(i.e.every month or every few months).One-third(34%)rarely or never intentionally use AI.Three in five(59%)use AI at least semi-regularly for personal purposes,with those not working or studying much less likely to use AI(only 37%).Three in five(58%)people who wor

179、k intentionally use AI regularly for work purposes,while four in five(83%)students regularly use AI in their studies.This high level of adoption reflects the ease with which AI systemsparticularly general-purpose generative AI toolscan be accessed and used by a diverse range of people and applied to

180、 a broad variety of tasks.This sets AI apart from many other advanced technologies that have greater barriers and constraints on access and use by individuals.38%of people report using AI on a weekly or daily basis.Trust,attitudes and use of AI:A global study 2025|19 2025 The University of Melbourne

181、.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.There are notable differences across countries in peoples adoption of AI,with emerging economies leading the wayThere is a distinct pattern of findin

182、gs between countries with advanced and emerging economies,with the use of AI tools notably higher in countries with emerging economies.On average,four in five(80%)people in emerging economies intentionally use AI tools on a regular or semi-regular basis,compared to three in five(58%)in advanced econ

183、omies.As shown in Figure 2,levels of AI use in most emerging economies exceed 70 percent of the population,with India and Nigeria reporting the highest regular or semi-regular usage(92%).Two emerging economies located in Eastern EuropeHungary and Romaniahave notably lower AI use compared to the othe

184、r emerging economies.In contrast,AI use levels in most advanced economies fall below 70 percent of the population,with the lowest usage reported in the Netherlands(43%)and the highest in Singapore(73%).Figure 1:Frequency of intentional use of AI tools for personal,work,or study purposes%Overall AI u

185、seIn your personal life(work/studies),how often do you intentionally use AI tools,including generative AI tools?Daily=most days or multiple times a day201415131721NeverFew times a yearEvery few monthsMonthlyWeeklyDailyTrust,attitudes and use of AI:A global study 2025|20 2025 The University of Melbou

186、rne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.669292919089888783828177757473727271706969686665636261616060575655535353525151505050505049494743OverallIndiaNigeriaUnited Arab EmiratesEgyptChinaS

187、audi ArabiaCosta RicaSouth AfricaBrazilTrkiyeMexicoArgentinaColombiaSingaporeChileLatviaNorwayEstoniaLithuaniaPolandIsraelSloveniaSwitzerlandPortugalRomaniaGreeceKoreaItalySpainDenmarkIrelandFinlandAustriaSlovak RepublicUnited States of AmericaUnited KingdomFranceGermanyAustraliaCanadaHungaryJapanNe

188、w ZealandCzech RepublicSwedenBelgiumNetherlands%Using AI on a semi-regular or regular basis:every few months,monthly,weekly or dailyFigure 2:Regular use of AI systems across countries%Emerging Economy%Advanced Economy66929291908988878382817775747372727170696968666563626161606057565553535352515150505

189、050504949474366Trust,attitudes and use of AI:A global study 2025|21 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.Most people have no AI training and half dont fee

190、l they understand AI,yet 3 in 5 believe they can use AI effectivelyDespite high levels of adoption,the majority of people report they have not received any form of AI training or education.Only two in five(39%)report some form of AI training,such as work-based AI training,formal or informal AI train

191、ing outside of work,or completing a university-level course related to AI(such as computer science or data analytics;see Figure 3).In line with these low levels of AI training,almost half(48%)report limited knowledge about AI,indicating that they do not feel they understand AI nor when or how it is

192、used.18 As shown in Figure 4,only one in five people report high levels of knowledge,and about a third report a moderate level.Despite low levels of AI education,training and knowledge,60 percent of people believe they can use AI effectively.This includes their ability to choose,use and communicate

193、with AI systems to support everyday activities,and evaluate the accuracy of AI output(see Figure 5).This is likely because many AI tools and systems are designed to be intuitive to use and accessible to a broad range of people(via a mobile phone application,for example,and by using natural language

194、to make requests),enabling these tools to be used widely with limited or no training.For example,AI voice assistants can be used simply by conversing with these tools.Figure 3:AI-related training or education%AI training%No AI training3961Figure 4:Self-reported AI knowledge%Low%Moderate%High483121To

195、 what extent do you.(a)Feel you know about AI?(b)Feel informed about how AI is used?(c)Think you understand when AI is being used?(d)Feel you have the skills and knowledge necessary to use AI tools appropriately?%Low=Not at all or To a small extent%High=To a large extent or To a very large extent Fi

196、gure 5:Self-reported AI efficacy%Disagree%Neutral%AgreeTo what extent do you agree with the following?I can%Disagree=Strongly disagree,Disagree,Somewhat disagree%Agree=Somewhat agree,Agree,Strongly agree242155Evaluate the accuracy of AI responses 231859Choose the most appropriate AI tool for a task

197、211960Communicate effectively with AI applications211762Skillfully use AI applications to help with daily workor activities211960AI efficacy overall Trust,attitudes and use of AI:A global study 2025|22 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International ent

198、ities.KPMG International entities provide no services to clients.All rights reserved.AI training,knowledge,and efficacy are lowest in the advanced economiesIn line with the distinct differences in the use of AI across economic groups,there are also pronounced differences between advanced and emergin

199、g economies when it comes to levels of AI training,knowledge,and efficacy.As shown in Figure 6,half of the people surveyed in emerging economies report having completed AI-related training or education,compared to less than a third in advanced economies.Similarly,almost two-thirds of people in emerg

200、ing economies report moderate or high knowledge about AI,compared to less than half in advanced economies.Around three-quarters of those in the emerging economies feel they can use AI tools and systems effectively,compared to only half in advanced economies.%Global%Emerging Economy%Advanced Economy3

201、95260324651506474AI trainingAI knowledgeAI efficacyFigure 6:AI training,knowledge and AI efficacy across economic groups As shown in Figure 7,AI training,knowledge,and efficacy are particularly high in Nigeria,Egypt,the UAE,India,China and Saudi Arabia.These six countries also rate highest on AI use

202、(see Figure 2).In contrast,AI training and knowledge are particularly low in Germany,the Czech Republic and Japan.Trust,attitudes and use of AI:A global study 2025|23 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities p

203、rovide no services to clients.All rights reserved.25%30%35%40%45%50%55%60%65%70%75%80%85%20%AI knowledge=%To a moderate extent,%To a large extent,%To a very large extent%AI efficacy=%Somewhat agree,%Agree,%Strongly agree%AI training=%Selected University level course in AI,%Selected Work-based traini

204、ng,or%Selected Formal or informal training outside workBolding indicates countries with emerging economies.Ordered by AI training.Figure 7:AI knowledge,efficacy,and training across countriesAI knowledgeAI efficacyAI trainingNigeriaEgyptUnited Arab EmiratesIndiaChinaSaudi ArabiaCost RicaSouth AfricaC

205、olombiaLithuaniaArgentinaBrazilMexicoEstoniaSwitzerlandSingaporeSloveniaChileNorwayIsraelSpainLatviaKoreaGreeceTrkiyeItalyDenmarkRomaniaPortugalIrelandFinlandPolandAustriaUSAUnited KingdomSlovak RepublicSwedenNew ZealandNetherlandsFranceCanadaBelgiumAustraliaJapanCzech RepublicGermanyHungaryTrust,at

206、titudes and use of AI:A global study 2025|24 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.A third are unaware that AI enables common applications they use:half do

207、nt know AI is used in social mediaAs an indicator of peoples objective awareness of AI use,respondents were asked if they use the three common technologies shown in Figure 8,and whether these technologies are enabled by AI(i.e.whether these technologies rely on AI to function).Seventy-nine percent o

208、f people use these common AI-enabled technologieshighlighting the prevalence of AI technologies in peoples livesbut over a third(36%)are unaware that these technologies use AI.Use of the technology does not necessarily translate into an increased understanding of whether AI is part of it.For example

209、,while the Self-reported understanding of AI has not changed over time and many are still unaware that AI is used in common applications like social media Despite the rapid uptake of AI since 2022,there has been no overall substantive change in self-reported knowledge of AI(M=2.6 in 2022;M=2.6 in 20

210、24).However,increases were found in four countries,Estonia,Brazil,China and South Africa,with the largest increases in Estonia(26%vs.50%,M=2.1 vs.2.8)and Brazil(38%vs.63%,M=2.5 vs.3.0).Although use of AI in common technologies such as social media,facial recognition,and virtual assistants has tended

211、 to remain constant or increased in most countries,many are still unaware that these technologies rely on AI to function.For example,social media use has remained constant and high over time across countries(88%use at both time points),yet many are still unaware that AI is used in social media platf

212、orms(2022:44%vs.2024:46%).Figure 8:Use of common technologies and awareness that they involve AI%Unaware this technology uses AI%Who use this technologyFor each technology below,please indicate if you have used it and if it uses AI7769792238369047Social mediaVirtualassistantsOverallFacialrecognition

213、majority(90%)of the sample reports using social media,nearly half(47%)of all respondents are unaware of AIs role in social media.As shown in Figure 8,this pattern of using technology without realizing it relies on AI is particularly strong for social media,but also evident in facial recognition and

214、virtual assistantsprompting the question of whether the awareness of AIs central role in these technologies would change how people engage with them.People in emerging economies are more likely to be aware that AI is used in these technologies than those in advanced countries(70%vs.61%),and they are

215、 also more likely to use these common AI-enabled technologies(88%vs.74%).Trust,attitudes and use of AI:A global study 2025|25 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights

216、reserved.Four in five want to learn more about AI,with interest highest in emerging economiesMost people(83%)are interested in learning more about AI,ranging from almost all(97%)in Nigeria to three in five(59%)in Australia.In most emerging economies,over 90 percent of people express a desire to lear

217、n more about AI.In contrast,respondents in seven advanced economies(Australia,New Zealand,the USA,Canada,the UK,Japan and Finland)have considerably lower interest(ranging from 59-67%),compared to other countries.Australia and Finland are notably low,with two in five(41%)people reporting no or low in

218、terest in learning more about AI.In summaryTaken together,these findings indicate high rates of AI adoption by the public,coupled with comparably low levels of AI training and literacy.Low levels of AI literacy may limit peoples ability to recognize the capabilities and applications of AI and thus f

219、ully realize benefits,and importantly,the ability to recognize the limitations of AI systems,critically evaluate their outputs,and guard against harm.For instance,social media users that are unaware of how algorithms shape content may fail to question the credibility or biases of algorithmically cur

220、ated content and face increased vulnerability to misinformation and manipulation.The findings also reveal accelerated uptake of AI tools and higher levels of AI literacy amongst people in emerging economies compared to advanced economies.This may be explained in part by the increasingly important ro

221、le that emerging and transformative technologies play in the economic development of these countries.19 As discussed in the next sections,people in emerging economies also tend to be more trusting,accepting,and positive about AI and experience the most benefits from its use,compared to those in adva

222、nced economies.In most emerging economies,over 90%of people express a desire to learn more about AI People with AI knowledge and efficacy tend to be more interested in learning more about AI(r=.48),suggesting a virtuous cycle where those who are already knowledgeable and confident in using AI are mo

223、re eager to learn and thus more likely to deepen their understanding further.In contrast,those with low knowledge and efficacy may fall further behind.Trust,attitudes and use of AI:A global study 2025|26 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International e

224、ntities.KPMG International entities provide no services to clients.All rights reserved.To what extent do people trust and accept AI systems?To answer this question,respondents were asked about their trust and acceptance of a range of AI systems,and the extent to which they perceive them to be trustw

225、orthy.They were also asked about the emotions they feel when it comes to AI applications.Our approach to measuring trust in AI aligns with the following common definition of trust:a willingness to be vulnerable to an AI system(e.g.by relying on system recommendations or output or sharing personal da

226、ta)based on positive expectations of how the system will operate(such as accuracy,helpfulness,data privacy and security).20 People have more trust in the technical ability of AI systems to provide a helpful service but are more skeptical of its safety,security and impact on peopleWhile most people u

227、se AI tools,many people have reservations about the trustworthiness of AI systems and their use in society.On average,58 percent of people view AI systems as trustworthy.21 People have more faith in the technical ability of AI systems to provide accurate and reliable output and services(65%)than in

228、their safety,security,impact on people,and ethical soundness(e.g.that they are fair,do no harm,and uphold privacy rights;52%).This difference is consistent across countries,as shown in Figure 9.To illustrate,in Finlanda country where trustworthiness is very lowhalf of the respondents view AI systems

229、 as providing a helpful service,yet only a third agree that these systems are safe and secure to use.By contrast,in Egyptwhere AI is perceived as highly trustworthy83 percent believe AI systems are accurate and provide a helpful service,while 72 percent agree that they are safe and secure to use.Tru

230、st is important because it underpins the acceptance and sustained adoption of AI.This is confirmed by our research:trust is associated with the acceptance and approval of AI systems(r=.70)and the use of AI(r=.48).People who trust AI systems are more likely to use them frequently.How trust in AI was

231、measuredTo understand how people view the trustworthiness of AI systems,we asked about two key components:the technical ability of AI(e.g.to provide accurate and reliable output and a helpful service),and safe and ethical use(e.g.to be safe and secure to use and ethically sound).We also examined two

232、 primary ways people demonstrate trust in AI systems:Reliance assesses peoples willingness to rely on an AI systems output,such as a recommendation or decision(i.e.to trust that it is accurate).Information sharing relates to the willingness to share information or data with an AI system(e.g.to provi

233、de personal information to enable the system to work or perform a service).Trust,attitudes and use of AI:A global study 2025|27 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All right

234、s reserved.40%45%50%55%60%65%70%75%80%85%90%35%Agree=Somewhat agree,Agree,Strongly agree.Ordered by perceived trustworthiness.Bolding indicates countries with emerging economies.Figure 9:Perceptions of the trustworthiness of AI systemsPerceived trustworthinessAbilitySafe and ethical useNigeriaIndiaC

235、hinaEgyptTrkiyeUnited Arab EmiratesSaudi ArabiaSouth AfricaBrazilCosta RicaMexicoSingaporeRomaniaChileColombiaSpainNorwayHungaryLithuaniaLatviaArgentinaItalyKoreaSwitzerlandPolandEstoniaPortugalSloveniaGreeceUnited KingdomIsraelCzech RepublicJapanIrelandDenmarkUSABelgiumAustriaGermanySlovak Republic

236、FranceAustraliaCanadaSwedenNew ZealandNetherlandsFinlandTrust,attitudes and use of AI:A global study 2025|28 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.Most peo

237、ple are ambivalent or unwilling to trust AI systems but accept their useThe concern about the safety and security of AI and its impact on people helps explain why a little over half(54%)of people are wary about trusting AI systems,reporting either ambivalence or an unwillingness to trust(see Figure

238、10).Only 46 percent are willing to trust AI systems.As peoples trust in AI may vary depending on the application of AI,we asked about trust in different AI use cases.As shown in Figure 10,there are similar levels of trust in generative AI tools,AI use in Human Resources,and AI systems in general(42-

239、45%are willing to trust,Ms=3.9-4.0).One difference is that people are more trusting of AI use in healthcare(52%willing,M=4.3),with healthcare the most trusted application in 42 of the 47 countries surveyed(see Figure 11).This difference likely reflects the direct benefit that increased precision of

240、medical diagnoses and treatments affords people,combined with generally high levels of trust in medical professionals in most countries.22 These findings reinforce that peoples trust of AI systems is contextual and can depend on the use case and their confidence in the organization that is deploying

241、 the AI system.Most people report low or moderate acceptance and approval of the use of AI systems(see Figure 10),with moderate acceptance indicating a level of ambivalence in their acceptance of AI use.In contrast,a third report high acceptance and approval.Taken together,these findings show that t

242、he majority(72%)have at least some level of acceptance of AI use.351946301852Healthcare AI391942Human Resources AI371944Generative AI361945AI in generalTrust in AI overall%Unwilling to trust%Ambivalent%Willing to trust%Unwilling to trust=Somewhat unwilling,Unwilling,or Completely Unwilling%Ambivalen

243、t=Neither willing nor unwilling%Willing to trust=Somewhat willing,Willing,or Completely willingFigure 10:Trust and acceptance of AI systemsHow willing are you to trust AI specific application?%Low acceptance=Not at all or Slightly%High acceptance=Highly or Completely2833AcceptanceTo what extent do y

244、ou accept/approve the use of AI specific application?%Low acceptance%Moderate%High acceptance39Trust,attitudes and use of AI:A global study 2025|29 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services

245、to clients.All rights reserved.25%30%35%40%45%50%55%60%65%70%75%80%20%Willing to trust based on Somewhat willing,Mostly willing and Completely willing.Ordered by%Willing.Bolding indicates countries with emerging economies.Figure 11:Trust in AI applications across countriesAIHuman Resources AIGenerat

246、ive AIHealthcare AlNigeriaIndiaEgyptChinaUnited Arab EmiratesSaudi ArabiaSouth AfricaTrkiyeBrazilHungaryNorwayCosta RicaSpainIsraelMexicoSingaporeLatviaSwitzerlandGreeceEstoniaArgentinaRomaniaColombiaChileKoreaUnited KingdomDenmarkPolandUSAItalyAustriaSloveniaIrelandPortugalSwedenAustraliaSlovak Rep

247、ublicBelgiumLithuaniaNew ZealandCanadaNetherlandsFranceGermanyCzech RepublicJapanFinlandTrust,attitudes and use of AI:A global study 2025|30 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clie

248、nts.All rights reserved.Trust and acceptance of AI is lower in advanced economiesAs shown in Figure 12,trust and acceptance of AI systems are consistently lower in advanced economies compared to emerging economies.In advanced economies,two in five are willing to trust AI systems by relying on their

249、output and sharing information with these systems.Half view AI systems as trustworthy,and two-thirds report at least moderate levels of acceptance.In contrast,people in emerging economies have more trust in AI systems,view them as more trustworthy,and have higher levels of acceptance and approval of

250、 their use.It is notable,however,that 43 percent of people in emerging economies remain ambivalent or unwilling to trust AI systems,highlighting that trust cannot be taken for granted.To illustrate this distinction at the country level,as shown in Figure 13,over half of the people surveyed trust AI

251、systems in 12 of the 17 emerging economies(ranging from 41 percent in Poland to 79 percent in Nigeria).Trust and acceptance are particularly high in the six emerging economies of Nigeria,India,Egypt,China,the UAE,and Saudi Arabiawith over 60 percent of people willing to trust AI and at least 49 perc

252、ent reporting high acceptance.These countries also have the highest levels of AI use and AI literacy,as previously reported.In contrast,less than half trust AI systems in 25 of the 29 advanced economies.Of the advanced economies,trust is highest in Norway,23 Spain,Israel,and Singapore(all over 50 pe

253、rcent willing to trust).In contrast,Finland and Japan rate the lowest on trust(25-28%)while New Zealand and Australia(15-17%high acceptance)rank lowest on acceptance.The higher trust and acceptance of AI in emerging economies is reflected in the accelerated uptake of AI in these countries.24Trust=%S

254、omewhat willing,Mostly willing,Completely willingTrustworthy=%Somewhat agree,Agree,Strongly agree trustworthyAcceptance=%Moderately,Highly,Completely accept%Global%Emerging EconomyFigure 12:Trust and acceptance of AI systems across economic groups%Advanced Economy465872395265576984Trust in AI system

255、sView AI as trustworthyAcceptance of AITrust,attitudes and use of AI:A global study 2025|31 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.%Trust=Somewhat willing,M

256、ostly willing or Completely willing%High acceptance=Highly or CompletelyBolding indicates countries with emerging economies.%Trust%High AcceptanceFigure 13:Trust and acceptance of AI systems across countriesNigeriaIndiaEgyptChinaUnited Arab EmiratesSaudi ArabiaSouth AfricaTrkiyeBrazilNorwayHungaryCo

257、sta RicaSpainMexicoIsraelSingaporeLatviaEstoniaSwitzerlandGreeceArgentinaRomaniaColombiaChileKoreaUnited KingdomUSAPolandDenmarkSloveniaItalyAustriaIrelandSwedenSlovak RepublicPortugalAustraliaBelgiumNew ZealandLithuaniaCanadaNetherlandsFranceGermanyCzech RepublicJapanFinland666361695452494444434332

258、38383136363536323328222130332830262022252417211535191824292218202734297976716865626256555452515151504746464545454442414141404040383636363635343434333332312825464754Trust,attitudes and use of AI:A global study 2025|32 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG In

259、ternational entities.KPMG International entities provide no services to clients.All rights reserved.People have mixed emotions about AI:both optimism and worry prevailPeople feel a range of emotions about AI applications.As shown in Figure 14,the majority feel optimistic and excited,while also worri

260、ed demonstrating a degree of emotional ambivalence.People in emerging economies report more positive emotions toward AI and a clear divergence between positive and negative sentiment.Optimism and excitement are dominant emotions in emerging economies,experienced by 74-82 percent of people.Significan

261、tly fewer(56%)feel worried.In contrast,people in advanced economies feel both worried and optimistic in almost equal measure(61-64%),with just over half(51%)feeling excited.Each emotion was measured on a 5-point scale,with the above figure displaying%Moderate to High=Moderately,Very or ExtremelyFigu

262、re 14:Emotions associated with AI%Global%Emerging Economy%Advanced Economy686160616451825674OptimisticWorriedExcitedIn thinking about AI specific application,to what extent do you feelare excited.In contrast,over 80 percent of people in China feel optimistic and excited about AI applications,while o

263、nly 43 percent feel worried.At least half of respondents feel worried about AI in all but three countries,underscoring that worry about AI often coexists with optimism and excitement in many countries.Reinforcing this pattern,Figure 15 shows emotions about AI applications at the country level.People

264、 in many advanced economies feel more worried than optimistic or excited,whereas optimism and excitement dominate in most emerging economies.To illustrate,70 percent of people in Japan feel worried and only 37 percent Trust,attitudes and use of AI:A global study 2025|33 2025 The University of Melbou

265、rne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.40%45%50%55%60%65%70%75%80%85%90%35%Figure 15:Emotions toward AI across countries%Optimistic%Excited%Worried%based on:%Moderately,%Very and%Extrem

266、ely.Ordered by%optimistic.Bolding indicates countries with emerging economies.IndiaChinaNigeriaTrkiyeEgyptUnited Arab EmiratesSaudi ArabiaCosta RicaBrazilSouth AfricaRomaniaLithuaniaLatviaNorwayChileMexicoColombiaIsraelArgentinaPolandSingaporeSloveniaEstoniaKoreaSpainHungarySwitzerlandItalyGreecePor

267、tugalGermanySlovak RepublicFranceDenmarkAustriaIrelandCzech RepublicBelgiumSwedenUSAUnited KingdomCanadaJapanFinlandNetherlandsNew ZealandAustraliaTrust,attitudes and use of AI:A global study 2025|34 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entit

268、ies.KPMG International entities provide no services to clients.All rights reserved.Trust in AI systems has decreased over time and worry has increasedThe perceived trustworthiness of AI systems decreased over time from 63 percent of people viewing AI systems as trustworthy in 2022 to 56 percent in 2

269、024(M=4.8 vs.M=4.6;see Figure 16).This demonstrates that many are feeling less positive about the ability of AI systems to provide accurate and reliable output,and be safe,secure and ethical to use.Perceived trustworthiness decreased in 13 of the 17 countries,with the largest decreases in Israel(68%

270、to 52%)and South Africa(76%vs.62%).Similarly,peoples willingness to rely on AI systems decreased on average from 52 percent in 2022 to 43 percent in 2024(M=4.3 vs.M=4.0;see Figure 16),with decreases in 12 of the 17 countries.The largest decreases occurred in Japan(43%to 21%)and Brazil(67%to 53%).Thi

271、s likely reflects that with increased use and exposure to AI systems,people have become more aware of their capabilities and limitations,prompting a more considered reliance on these tools.Over this same period,there is a striking increase in the number of people feeling worried about AI systems,ris

272、ing from almost half(49%)of respondents in 2022 to 62 percent in 2024(M=2.4 to M=3.0).This increase was found in 15 of the 17 countries,with the largest increases in Brazil (49%in 2022 vs.75%in 2024)together with Israel,Estonia,the Netherlands and Finland(ranging from 21-26%increase in worry).In 11

273、of the 17 countries,people also feel less excited about AI systems,with the largest difference in France,where just 35 percent feel excited about AI in 2024(M=2.0)compared to 58 percent in 2022(M=2.6).The only country where excitement increased is Korea,where 75 percent report feeling excited in 202

274、4,compared to 57 percent in 2022(M=3.2 vs.M=2.5).Figure 16:Trust of AI systems and worry about AI in 2022 and 202452%43%63%56%49%62%0%10%20%30%40%50%60%70%80%20222024%Willing to rely on AI systems%Perceive AI systems as trustworthy%Worried about AI systemsWilling to rely on AI systems=Somewhat willi

275、ng,Willing,or Completely willingPerceived trustworthiness of AI systems=aggregate Somewhat agree,Agree,or Strongly agree Worried about AI systems=moderately,very,extremelyTrust,attitudes and use of AI:A global study 2025|35 2025 Copyright owned by one or more of the KPMG International entities.KPMG

276、International entities provide no services to clients.All rights reserved.2025 The University of Melbourne.In summaryOverall,the findings reveal considerable ambivalence toward the use of AI systems in society,stemming from the tension that people are less trusting of the safety and security of usin

277、g AI systems and their impact on society,but are more trusting of their technical ability to provide a helpful service.This tension is reflected in low and ambivalent trust of AI,moderate acceptance,and the coexistence of optimism with worry,particularly for people in advanced economies.Moreover,tru

278、st in AI has declined over time,while worry has increased.The next section examines how this ambivalent trust is shaped by perceptions and experiences of the benefits and risks of AI systems.Trust,attitudes and use of AI:A global study 2025|36 2025 The University of Melbourne.2025 Copyright owned by

279、 one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.How do people view and experience the benefits and risks of AI?To help answer this question,we asked the extent to which people perceive and have observed or experienced ben

280、eficial or negative outcomes from AI,and if they feel the benefits of AI applications outweigh the risks.People expect and are experiencing a broad range of benefits from AI Most people(83%)believe the use of AI will result in a wide range of benefits,as shown in Figure 17.Importantly,73 percent of

281、people are personally experiencing or observing these benefits.25The most commonly expected benefits are also some of the most realized,with over three quarters reporting they have experienced or observed improved efficiency and effectiveness,reduced time spent on mundane or repetitive tasks and imp

282、roved levels of accessibility to information or services.26 Increased fairness due to the use of AI(e.g.by reducing human bias)is the least commonly realized benefit,but it is still experienced or observed by over half of respondents(54%).The utility of AI and peoples lived experience of its benefit

283、s help explain the widespread use,adoption and qualified acceptance of AI technologies,despite the trust concerns.The positive benefits experienced are largely performance orientedin line with our finding that people are more trusting of AIs ability to provide a helpful service and output.People who

284、 expect and experience or observe benefits from AI are more likely to trust(r=.42-.57),accept(r=.41-.63),and use AI(r=.40-.41).They are also more likely to have AI training or education(r=.25),AI knowledge(r=.31-.38),and AI efficacy(r=.38-.45).73%are personally experiencing or observing benefits of

285、AI.Trust,attitudes and use of AI:A global study 2025|37 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.81827878726872747173695473Figure 17:Expected and experienced

286、benefits of AI use%Low%Moderate to High%Personally experienced or observedI expect the use of AI specific application will result in these potential positive outcomes1711121415161617181819252683898886858484838282817574Overall benefits Reduced time spent on mundane orrepetitive tasksImproved efficien

287、cyImproved accessibilityImproved effectivenessEnhanced precision or personalizationReduced costs or better use of resourcesEnhancing what people can doEnhanced decision-making andproblem-solvingImproved outcomes for peopleEnhanced creativityIncreased fairness%Low=Not at all or To a small extent%Mode

288、rate to High=To a moderate extent,To a large extent or To a very large extentInnovationPeople in emerging economies are more likely to expect and realize the benefits of AINinety percent of people in emerging economies expect benefits from AI applications,compared to 79 percent in advanced economies

289、.As shown in Figure 18,people in emerging economies have the most positive expectations of the benefits of AI.For instance,95 percent of people in Nigeria expect a wide range of benefits.In contrast,fewer people expect benefits from AI in several advanced economies,particularly Australia,Canada,Finl

290、and,Japan,New Zealand,the UK and the USA.The majority of people in emerging economies are also more likely to have observed or experienced AI benefits(82%vs.65%in advanced economies).The largest differences between economies relate to the benefits of increased fairness(66%vs 43%),enhanced creativity

291、(80%vs 59%),and improved outcomes for people(84%vs 64%).AI systems may be perceived and experienced as more beneficial in emerging economies because of their ability to fill critical resource gaps and provide greater relative opportunities to people.For instance,the use of AI systems in healthcare h

292、as the potential to enhance service delivery and improve health outcomes in areas where there is limited access to medical professionals.Trust,attitudes and use of AI:A global study 2025|38 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG I

293、nternational entities provide no services to clients.All rights reserved.Figure 18:Expected benefits of AI across countriesReduced time spent on repetitive or mundane tasksImproved accessibility Improved efficiencyEnhanced precision or personalizationImproved effectivenessReduced costs or better use

294、 of resourcesEnhancing what people can doInnovationImproved outcomes for peopleEnhance decision-making or problem solvingIncreased fairnessEnhanced creativity45%50%55%60%65%70%75%80%85%90%95%100%40%NigeriaTrkiyeEgyptMexicoItalyIndiaChileUnited Arab EmiratesSaudi ArabiaIsraelGreeceCosta RicaColombiaC

295、hinaBrazilArgentinaSpainSouth AfricaPortugalPolandKoreaSingaporeRomaniaFranceDenmarkNorwayGermanySwedenLithuaniaEstoniaCzech RepublicBelgiumSloveniaSlovak RepublicNetherlandsHungaryAustriaSwitzerlandLatviaIrelandCanadaUSAUnited KingdomJapanNew ZealandAustraliaFinlandBased on%Moderate to High=To a mo

296、derate extent,To a large extent or To a very large extentOrdered by Reduced time spent on repetitive or mundane tasks.Bolding indicates countries with emerging economies.Trust,attitudes and use of AI:A global study 2025|39 2025 The University of Melbourne.2025 Copyright owned by one or more of the K

297、PMG International entities.KPMG International entities provide no services to clients.All rights reserved.Based on%Yes.Ordered by Reduced time spent on repetitive or mundane tasks.Bolding indicates countries with emerging economies.Figure 19:Experienced benefits of AI across countries30%35%40%45%50%

298、55%60%65%70%75%80%85%90%95%25%Reduced time spent on repetitive or mundane tasksImproved accessibility Improved efficiencyEnhanced precision or personalizationImproved effectivenessReduced costs or better use of resourcesEnhancing what people can doInnovationImproved outcomes for peopleEnhance decisi

299、on-making or problem solvingIncreased fairnessEnhanced creativityNigeriaEgyptChileUnited Arab EmiratesSaudi ArabiaSingaporeMexicoIndiaCosta RicaRomaniaColombiaGreeceChinaArgentinaTrkiyeBrazilSpainSouth AfricaIsraelItalyLithuaniaKoreaLatviaJapanGermanyPortugalEstoniaSwitzerlandSloveniaDenmarkPolandHu

300、ngarySlovak RepublicCanadaAustriaUnited KingdomNorwaySwedenIrelandFranceAustraliaUSACzech RepublicBelgiumNew ZealandFinlandNetherlandsTrust,attitudes and use of AI:A global study 2025|40 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG Inte

301、rnational entities provide no services to clients.All rights reserved.People are concerned about a range of negative outcomes from AI use and two in five are experiencing negative outcomesWhile many of those surveyed are experiencing significant benefits from AI use,the majority(79%)are also concern

302、ed about a broad range of risks and negative outcomes from AI use(see Figure 20).Many of these risks are at the societal level,impacting society broadly rather than having isolated impacts on the individuals who use AI.27 Cybersecurity risk(e.g.from hacking or malware)is a dominant concern raised by

303、 85 percent of people,together with the loss of human interaction and connection(e.g.losing the option to speak with a human service provider).Other risks raised by over 80 percent of people include misinformation and disinformation(e.g.AI used to spread misleading or false information and deepfakes

304、),manipulation or harmful use,loss of privacy or intellectual property(IP),deskilling and dependency,and job loss.In comparison,people are less concerned about the risk of bias or unfair treatment from AI use or the environmental impact(68-69%).This may reflect a lack of awareness of the potential f

305、or AI systems to codify existing biases in datasets,and the high energy usage required to develop some AI systems and power the data centers they rely on.Although the percentages are lower,bias and environmental impact remain clear concerns for more than two thirds of people.In addition to being con

306、cerned about the risks of AI applications,two in five have personally experienced or observed these negative outcomes(43%;see Figure 20).The loss of human interaction and connection,inaccurate outcomes,and misinformation and disinformation are the most commonly experienced negative outcomes from AI(

307、52-55%).Bias or unfair treatment is the least commonly experienced or observed outcome,but it was still experienced by almost a third of people.Figure 20:Perceived risks and experienced negative outcomes from AI use%Low%Moderate to High%Personally experienced or observedHow concerned are you about t

308、hese potential negative outcomes of AI specific application?%Low=Not at all or To a small extent%Moderate to High=To a moderate extent,To a large extent or To a very large extent 43Overall risks217944Cybersecurity risks158555Loss of human interaction and connection178352Misinformation or disinformat

309、ion188248Deskilling and dependency188241Loss of privacy or intellectual property188240Manipulation or harmful use198142Job loss208046System failure217934Human rights being undermined217954Inaccurate outcomes237740Disadvantage due to unequal access to AI2476Environmental impact34316931Bias or unfair

310、treatment3268Trust,attitudes and use of AI:A global study 2025|41 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.The risks of AI are viewed and experienced in a com

311、parable way across countriesIn contrast to the differences across countries in how people view the benefits of AI,there are few differences across countries in peoples concerns about the risks:the same proportion of people are concerned about negative outcomes from AI in both advanced and emerging e

312、conomies(79%and 78%,respectively)and the majority of people in all countries report moderate or high concern about these risks(ranging from 67%in China to 87%in Greece).As shown in Figure 21,the top concerns in almost all countries are either cybersecurity risks or the loss of human connection.China

313、,Egypt,Nigeria,Saudi Arabia and South Africa are the exceptions,where job loss is the primary or an equal concern.There are also commonalities in what people are least concerned about,with either the environmental impacts of AI or the potential risk of bias from AI ranking last in every country.The

314、experience or observation of negative outcomes is also similar across economies(Emerging:46%vs.Advanced:40%).However,as shown in Figure 22,there is a trend for people in emerging economies to be more likely to have experienced or observed job loss due to AI (46%vs.34%in advanced economies).People in

315、 emerging economies are more likely to believe the benefits of AI outweigh the risks:opinion is divided in advanced countriesGlobally,42 percent of people believe the benefits of AI outweigh the risks,compared to 32 percent who believe the risks outweigh the benefits,and 26 percent who believe benef

316、its and risks are balanced.This aligns with the finding that more people report experiencing benefits from AI than negative outcomes.However,there are significant country differences in how people perceive the balance between AI risks and benefits.Half of people in emerging economies believe benefit

317、s outweigh risks,but opinions are more divided in advanced economies,where 38 percent believe the benefits outweigh risks and an almost equal number(37%)believe the risks outweigh the benefits.This aligns with the previously reported finding that more people in emerging economies expect and experien

318、ce benefits from AI.As shown in Figure 23,over 60 percent believe benefits outweigh risks in Nigeria,China,and Egypt(from 61%in Egypt to 74%in Nigeria).In contrast,a third or less agree that the benefits outweigh the risks in Australia,New Zealand,the Netherlands,Sweden,Finland,Canada,Ireland,and Fr

319、ance.Although perspectives on AI vary across economies,in no country does the belief that AI risks outweigh the benefits reach 50 percent.This suggests that,despite concerns,most people in all countries acknowledge the benefits of AI systems.Trust,attitudes and use of AI:A global study 2025|42 2025

320、The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.Figure 21:Concerns about the risks of AI across countriesCybersecurity risksMisinformation or disinformationLoss of human

321、interaction and connectionLoss of privacy or intellectual propertyDeskilling and dependencyManipulation or harmful useSystem failureJob lossInaccurate outcomesBias or unfair treatment Human rights being undermined Environmental impactDisadvantage due to unequal access to Al 55%60%65%70%75%80%85%90%B

322、ased on%Moderate to High=To a moderate extent,To a large extent or To a very large extentOrdered by%Cybersecurity risks.Bolding indicates countries with emerging economiesPortugalGreeceSpainNetherlandsMexicoSwedenSingaporePolandItalyDenmarkCzech RepublicColombiaArgentinaSouth AfricaSlovak RepublicIr

323、elandGermanyChileBelgiumTrkiyeKoreaFranceCanadaUnited KingdomRomaniaIsraelBrazilAustraliaSloveniaNew ZealandFinlandCosta RicaIndiaHungaryUSANorwayEstoniaLithuaniaAustriaUnited Arab EmiratesSwitzerlandLatviaJapanSaudi ArabiaNigeriaEgyptChina Trust,attitudes and use of AI:A global study 2025|43 2025 T

324、he University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.Figure 22:Experienced negative outcomes from AI use across countriesCybersecurity risksMisinformation or disinformationLoss

325、 of human interaction and connectionLoss of privacy or intellectual propertyDeskilling and dependencyManipulation or harmful useSystem failureJob lossInaccurate outcomesBias or unfair treatment Human rights being undermined Environmental impactDisadvantage due to unequal access to Al Based on%Yes.Or

326、dered by Cybersecurity risks.Bolding indicates countries with emerging economies20%25%35%30%45%40%50%55%60%65%LatviaIndiaSingaporeUnited Arab EmiratesColombiaSaudi ArabiaRomaniaChinaTrkiyeEstoniaEgyptCosta RicaChileNigeriaMexicoArgentinaGreeceSwitzerlandFinlandLithuaniaSloveniaSlovak RepublicDenmark

327、BrazilBelgiumAustriaIrelandSouth AfricaKoreaSpainNorwayAustraliaUSAPortugalJapanIsraelHungaryCzech RepublicSwedenFinlandFranceNetherlandsCanadaItalyUnited KingdomGermanyNew ZealandTrust,attitudes and use of AI:A global study 2025|44 2025 The University of Melbourne.2025 Copyright owned by one or mor

328、e of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.For you personally,how do the benefits of AI specific application compare to the risks?%Benefits outweigh risks=Benefits slightly outweigh the risks,Benefits outweigh the risks,and Ben

329、efits strongly outweigh the risks.Light bars and bolding indicate countries with emerging economiesFigure 23:Perceptions across countries that AI benefits outweigh risks427469615555535151494847464544444444434242424241414141404039393837373737353535343333323232313130OverallNigeriaChinaEgyptIndiaSaudi

330、ArabiaUnited Arab EmiratesSouth AfricaTrkiyeKoreaCosta RicaSingaporeLithuaniaNorwayArgentinaBrazilChileSpainSwitzerlandIsraelLatviaMexicoPolandEstoniaItalyJapanRomaniaDenmarkSlovak RepublicColombiaGreeceSloveniaCzech RepublicGermanyHungaryUnited KingdomAustriaPortugalUSABelgiumFranceIrelandCanadaFin

331、landSwedenNetherlandsNew ZealandAustraliaTrust,attitudes and use of AI:A global study 2025|45 2025 The University of Melbourne.2025 Copyright owned by one or more of the KPMG International entities.KPMG International entities provide no services to clients.All rights reserved.In summaryTaken togethe

332、r,the extensive range of benefits and negative outcomes experienced from AI use highlights the paradoxical impacts of AI systems on individuals and society.For example,depending on how it is implemented and for what purpose,AI systems can either increase fairness or augment bias,facilitate accurate

333、information or contribute to misinformation,enhance what people can do or deskill people.As with all powerful technologies that augment capabilities and offer transformative opportunities for advancement and growth while also augmenting risks and negative outcomes,AI systems require careful management and governance,together with guardrails and guidance to ensure appropriate and responsible use an