People · Governance · 6 min read

Your staff are already using AI. Half of them are hiding it.

In Australia, 51% of employees have passed off AI output as their own and 44% have used it against policy. Only a quarter have had any training. The fix is not a ban.

Team level only · n ≥ 5 · band shown on every score · never an individual verdict

The short version

KPMG and the University of Melbourne's 2025 global study found 38% of Australian employees intentionally use AI at work. Of the workforce, 44% report using AI in ways that breach workplace policy and 51% report presenting AI-generated content as their own. Only 36% of Australians are willing to trust AI systems and 24% report any formal or informal AI training. Australia had the lowest share of any surveyed country believing AI's benefits outweigh its risks, at 30%.

What your staff are doing already

Use is ahead of policy, training and trust.

0%30%60%Use AI at work38%Use it against policy44%Pass AI work off as own51%Willing to trust AI36%Have had any training24%
Signal bars are behaviour, ink bars are the supports that should sit underneath it. Australia recorded the lowest share of any surveyed country believing AI's benefits outweigh its risks (30%).

Source · KPMG and University of Melbourne, Trust in AI global study, 2025, 2025

Why they hide it

People hide what they think will be punished or will make them look replaceable. Gallup's US tracking shows regular use climbing every quarter (30% weekly by May 2026) while only 25% of employees say their organisation has communicated a clear plan for integrating AI. Microsoft's 2025 Work Trend Index found 33% of leaders considering headcount reductions because of automation, and 83% saying AI will let employees take on more strategic work earlier. Your people have read both of those numbers. Silence from you means they assume the first one.

The three things that stop shadow use

A plain policy that says what is allowed, in one page. Training that is measured, not just offered. And an honest read of who on the team is adapting well and who is anxious, so the manager conversation happens before the mistake does. None of that is expensive. All of it is rarer than it should be.

The regulatory clock

The Privacy and Other Legislation Amendment Act 2024 brings transparency obligations for automated decision making that affects individuals, with the provisions commencing in December 2026. The EU AI Act classifies recruitment and worker management systems as high risk. If AI touches hiring, rostering, performance or pay in your business, "we did not know staff were using it" will not be an answer.

Common questions

How many Australian employees use AI against workplace policy?
44%, and 51% have presented AI-generated content as their own, according to the KPMG and University of Melbourne global trust study (2025).
How many employees have had AI training?
24% of Australians report any formal or informal AI training (KPMG and University of Melbourne, 2025). Only 25% of employees say their organisation has a clear AI plan (Gallup, 2026).
What should an AI policy for employees include?
A one page statement of what is allowed, measured training rather than offered training, and a team level read of who is adapting and who is anxious before the mistake happens.

Sources · KPMG and University of Melbourne, Trust, attitudes and use of AI, 2025 (Australian snapshot). Gallup, AI at work indicator, May 2026. Microsoft, Work Trend Index 2025. Privacy and Other Legislation Amendment Act 2024.

Related