Survey Assessing Risks from AI 2025
Understanding Australian public views on AI risks and governance
We conducted a representative survey of 933 Australian adults to understand public perceptions of AI risks and support for AI governance actions in Australia.
Key Insights
Comparing surveys: View 2024 findings
2025 Survey (933 Australians): Australians expect AI to be as safe as commercial aviation - at least 4,000x safer than current risk estimates. They want the government to better manage AI risks, and many risk controls would increase their trust in AI.
What This Means for Australia
Safety standards matter: Australians hold AI to the same rigorous safety standards as commercial aviation. This means they expect AI systems to cause fewer than 1 death per 100 million hours of operation—far stricter than current AI safety benchmarks.
Public trust depends on action: The research shows that implementing appropriate risk controls would directly increase trust in AI technology. This presents a clear pathway for both government and industry.
Governance framework needed: With strong public support for government action on AI risks, there is a mandate for developing comprehensive AI governance frameworks that address current harms and potential catastrophic risks.
How to Read This Research
- 2025 Technical Report (interactive report) - Complete findings, read online
- 2025 Technical Report (PDF) - Complete report for download and offline reading
Suggested citation: Noetel, M., Saeri, A.K., Graham, J., & Slattery, P. (2025). Survey Assessing Risks from Artificial Intelligence: 2025 Technical Report. aigovernance.org.au
How We Conducted This Survey
We analysed data from 933 Australians, recruited through online representative quota sampling stratified by age, sex, and Australian state / territory. We also conducted multilevel regression with poststratification to construct more accurate population estimates based on 2021 Australian Census data.
This project is a collaboration between Ready Research and The University of Queensland. The project team includes Dr Alexander Saeri, Dr Michael Noetel, Jessica Graham, and Dr Peter Slattery.
