Australia is at or near the bottom globally on AI trust and optimism. In a 47-country study, only 36% of Australians say they’re willing to trust AI, even though half use it regularly, and 78% worry about negative outcomes. Most starkly, just 30% say AI’s benefits outweigh the risks – the lowest share of any country surveyed.
Put simply: Australians want stronger guardrails and clearer accountability before they lean in.
Reference: University of Melbourne & KPMG (2025). Trust, Attitudes and Use of Artificial Intelligence: Australia insights. https://assets.kpmg.com/content/dam/kpmg/au/pdf/2025/trust-in-ai-global-insights-2025-australia-snapshot.pdf
What the data shows (Australia & global)
-
Australia’s trust gap: University of Melbourne & KPMG (2025) report only ~36% willing to trust AI, despite ~50% using it regularly. 78% worry about negative outcomes; 77% support regulation.
-
Workers & consumers want accountability: UTS Human Technology Institute (2025) finds low understanding and trust, and clear expectations for transparency, human oversight and avenues for redress.
-
Global trend: Stanford’s 2025 AI Index shows cautious optimism in many countries but persistent concerns around safety, fairness, and company data practices.
-
Jobs anxiety persists: ANU polling indicates many Australians expect more job losses than gains from AI/automation.
What Australian government expects (now)
-
Voluntary AI Safety Standard (DISR): 10 guardrails covering risk management, human oversight, transparency/contestability, and incident response. Practical starting point for all orgs; likely basis for future mandatory rules in high‑risk settings.
-
OAIC privacy guidance: Do not enter personal or sensitive information into public GenAI tools; follow Australian Privacy Principles (APPs) when using or training AI (e.g., consent, purpose, notices, security, access and correction).
-
National framework for AI assurance in government: maps practices to Australia’s AI Ethics Principles; stresses lawful, ethical, human‑centred use and clear accountability.
-
Public‑sector guidance: DTA’s interim guidance for generative AI use; Commonwealth Ombudsman’s Automated Decision‑Making Better Practice Guide; National Archives guidance that AI‑generated content is a record and must be managed under the Archives Act.
-
National AI Centre (NAIC): programs like the Responsible AI Network (RAIN) to uplift practice across law, standards, governance, leadership, technology and design.
Global guardrails worth tracking
-
EU AI Act: phased application from 2025–2027; full enforcement from August 2026 for most obligations, with bans and transparency duties earlier. Expect global spill‑over via supply chains and vendor contracts.
-
NIST AI Risk Management Framework (US): widely used, plus a Generative AI Profile (2024) , helpful for risk registers, testing plans and controls.
-
ISO/IEC 42001: management‑system standard for AI (AIMS) , align your policies, roles, risk processes and continuous improvement to it.
What this means for Australian organisations
-
Make trust visible: explain where AI is used, provide human‑in‑the‑loop, and offer a clear path to challenge/appeal decisions.
-
Apply the Voluntary AI Safety Standard to each use case: document risks, controls, owners, test evidence and incident playbooks.
-
Lift literacy & policy: set a short AI policy; run role‑based training; label AI use for staff and customers.
-
Co‑design with people: involve workers and customers early to avoid ‘so‑so automation’ and adoption friction.
-
Measure outcomes and trust: track error/bias rates, appeals, security events and user sentiment alongside ROI.
What this means for AI service providers
-
Governance‑by‑design deliverables: risk register, evaluation & red‑team results, model cards, oversight design, transparency notes, and privacy impact assessments.
-
Map to standards: show alignment to VAISS guardrails, NIST AI RMF (incl. GenAI Profile), ISO/IEC 42001 and OECD/G7 principles.
-
Privacy first: follow OAIC guidance; avoid feeding personal or sensitive data into public GenAI; design data‑minimisation and purpose limits.
-
Change & skills: bundle deployments with short training, office hours and clear operating manuals.
-
Honesty on jobs impact: include role redesign and augmentation plans; don’t oversell automation.
How Agile Insights helps (fast, safe, assured)
-
AI Trust & Governance Workshops
-
Solution‑level Assurance Pack: testing & eval summary, risk register, model card, oversight plan, transparency notes, incident response.
-
People‑centred adoption: worker/customer co‑design, role‑based training
-
Executive briefings with Microsoft: shortlist high‑ROI use cases, start small, then scale what works.
References
University of Melbourne & KPMG (2025). Trust, Attitudes and Use of Artificial Intelligence: Australia insights. https://assets.kpmg.com/content/dam/kpmg/au/pdf/2025/trust-in-ai-global-insights-2025-australia-snapshot.pdf
UTS Human Technology Institute (2025). Disconnected AI: the unmet expectations of consumers and workers (Insight Summary). https://www.uts.edu.au/globalassets/sites/default/files/2025-03/HTI-Insight-Summary-Disconnected-AI.pdf
ANU Centre for Social Research & Methods (2023). Views of Australians towards science and AI (ANU poll) https://polis.cass.anu.edu.au/files/docs/2025/6/Views_of_Australians_towards_science_and_AI_-_For_web.pdf
Department of Industry, Science and Resources (2024/25). Voluntary AI Safety Standard. https://www.industry.gov.au/publications/voluntary-ai-safety-standard