Australian workers are secretly using Generated Artificial Intelligence (GEN AI) tools without knowledge or approval from their superiors, according to a new report.
The “Movement of Our GEN AI: Impact on Jobs and Skills” report from federal government jobs and skills points to several studies in Australia, showing that between 21% and 27% of workers (particularly the white-collar industry) use AI behind the back of managers' backs.
Why do some people still hide it? The report says people have generally said:
- “I feel that using AI is fraudulent.”
- Has a “fear of being considered lazy”
- And “they are feared that they are considered to be under-capacity.”
Most impressive is the rise in unapproved “shadow use” of AI, despite the Federal Accounting and Productivity Commission urging Australians to make the most of AI.
The results of the new report highlight the gaps in how AI is used in the workplace, highlighting that workers and employers are doing the right thing in the dark.
As I've seen in my work, as a legal researcher looking at AI governance and as a practical lawyer – there are some jobs that change as soon as the rules for using AI in the workplace cross the borders of states within Australia.
ai's risks and benefits of using shadows
The 124-page Employment and Skills Australia Report covers many issues, including early and uneven recruitment of AI, how AI can help future jobs, and how it affects employment availability.
Among its most interesting findings, it relates to workers who use AI secretly, but this is not always a bad thing. The report found that people using AI in the shadows can sometimes be “promoting bottom-up innovation in some sectors,” a hidden leader.
However, there are also serious risks.
Worker-led “shadow use” is an important part of previous recruitment. A significant portion of employees often use GEN AI tools independently without employer oversight, showing grassroots enthusiasm, as well as raising concerns about governance and risk.
This report recommends taking advantage of this early adoption and experiment, but warns.
In the absence of clear governance, shadow use can multiply. This informal experiment, although a source of innovation, can also fragment practices that are difficult to expand or integrate later. It also increases risks regarding data security, accountability and compliance, and inconsistent outcomes.
Actual risks from AI failures
The report calls for the national management of Australia's Gen AI to transition through coordinated national frameworks, focus capabilities, and a group-wide boost of digital and AI skills.
This reflects my own research, and there are blind spots in the legal framework of AI in Australia, and the knowledge system from law to legal reporting requires a fundamental rethink.
Even some professions where clearer rules have emerged too often come after serious failure.
In Victoria, child protection workers entered ChatGpt-sensitive details about lawsuits relating to sexual offences against young children. The Victorian intelligence committee bans state child protection staff from using AI tools until November 2026.
Lawyers have also been found to misuse AI, from the US, the UK to Australia.
Yet another example – containing misleading information created by AI for the Melbourne murder case – was reported yesterday.
But even for lawyers, rules are patchy and vary from state to state. (Federal courts are one of those still developing rules.)
For example, New South Wales lawyers are clearly not permitted to use AI to generate the contents of their affidavits.
However, other states and territories have not adopted this position clearly.
This article is part of a series of conversations about Jobs in the age of AI. Key experts look at what AI means for workers at different career stages, how AI is reshaping our economy, and what we can do to prepare.
Clearer rules as a workplace and as a nation
Currently, using AI in the workplace is in the Governance Grey Zone. Most organizations are carried out without clear policies, risk assessments, or legal protection measures. Even if everyone is doing it, the first one will face consequences.
In my view, a national unified AI law would be desirable. After all, the AI technology we use is the same whether you are in New South Wales or in the Northern Territory. AI doesn't know physical boundaries. But I don't think that's very likely yet.
What can they do if employers don't want workers to use AI secretly? If there is a clear risk, start by providing clear policies and training to workers.
One example is what legal professions do now (in some states). It's not perfect, but it's a step in the right direction.
But that's definitely not enough, especially since the rules aren't the same nationwide.
Clearly positioned positions for both workers and employers requires more aggressive national AI governance with clearer policies, training, ethical guidelines, a risk-based approach and compliance monitoring.
Without a national AI governance policy, employers are left to navigate fragmented, inconsistent regulatory minefields, seeking violations every turn.
Meanwhile, the very workers who may be at the forefront of our AI transformation may be driven to use AI secretly for fear of being judged as a lazy cheat.
