Government moves to rein in workplace AI

The Albanese government is edging towards a new regulatory framework for artificial intelligence in Australian workplaces – one that puts trust, safety and shared benefits at the centre, but stops short of giving unions a veto over how the technology is deployed.

Workplace Relations and Employment Minister Amanda Rishworth has used the AFR Workforce Summit to outline the government’s first concrete moves on AI at work, framing them as a balancing act between protecting workers and unlocking productivity.

On one side are unions warning that AI-driven restructures and job losses are already here. On the other are employers nervous that heavy-handed rules could smother innovation before the promised efficiency gains materialise.

Rishworth’s message to both camps is that regulation is coming – but it will be deliberative rather than knee-jerk.

She has commissioned a “gap analysis” to test whether existing workplace laws and institutions are equipped to deal with AI, or whether a new, overarching legislative framework is needed. At the same time, she is adamant about preserving workers’ access to justice, ruling out changes to the Fair Work Act’s no‑costs regime even as AI-generated unfair dismissal claims swamp the commission.

The centrepiece of the Albanese government’s AI-at-work strategy is the new AI Employment and Workplaces Forum, a tripartite body bringing together unions, major employer groups and government.

The forum, which meets for the first time in Adelaide this week, will convene at least three times a year and includes ACTU secretary Sally McManus alongside the chief executives of the Business Council of Australia, Australian Industry Group and the Australian Chamber of Commerce and Industry.

Its remit is broad: to “chart a tripartite path” on how AI is adopted, ensure productivity gains are shared with workers and test whether current laws are “fit for purpose”.

But Rishworth has been explicit about what the forum is not: it will not give unions the power to veto AI deployment in workplaces.

That line is clearly aimed at business groups anxious that consultation about AI could harden into an effective ban if unions object to specific technologies or restructuring plans. Unions, meanwhile, have pushed for measures such as six‑month notice periods for AI-related job cuts and stronger consultation duties, reflecting fears that hundreds of roles could disappear in the coming months.

Rishworth is frank that the forum will not deliver “a silver bullet solution”. Instead, she casts it as an exercise in building a common evidence base and shared language around AI so that future regulatory interventions rest on facts rather than hype.

AI’s labour market impact: disruption or augmentation?

Central to that evidence base is new data from Rishworth’s department on how AI has affected jobs since the launch of ChatGPT in late 2022. The report, due next month, looks at labour market changes up to February 2026, with a particular focus on entry-level roles and occupational mix.

Contrary to some of the more apocalyptic forecasts, the early data suggests the sky has not yet fallen. Employment outcomes for young tertiary graduates, often described as the “canaries in the coalmine” for white-collar automation, have remained positive. Nor is there an “elevated rate of compositional change” in the labour market – in other words, the overall mix of jobs in the economy is not shifting faster than usual.

However, the picture is not entirely benign. The government is detecting a “slight softening” in the rate of growth for occupations most exposed to AI, such as filing clerks and keyboard operators, hinting at gradual attrition in traditional administrative roles.

Rishworth is careful not to oversell the findings. The data is “reflective of a particular point in time, and is not predictive”, she cautions. It also predates some of the most high‑profile technology-related layoffs, including thousands of job cuts at Wisetech, Atlassian and Bendigo and Adelaide Bank, where AI and automation have been explicitly cited as drivers.

For now, Jobs and Skills Australia still sees augmentation – humans working alongside AI, rather than outright replacement – as the most likely medium-term scenario. But the government clearly wants to get ahead of any tipping point where that calculus changes.

Rishworth’s concern is not limited to headcount. She has repeatedly stressed that “work intensification” driven by AI may represent a more immediate hazard than outright job destruction.

As AI systems automate routine tasks and accelerate workflows, there is a risk employers simply ratchet up expectations rather than redesign jobs. The result can be longer hours, greater surveillance, and “always on” performance metrics that erode psychological safety.

Rishworth is candid that she does not yet have a complete policy blueprint for managing these risks. But with psychological injury claims already escalating, she is clear that AI-related intensification should be treated as a serious workplace health and safety issue, not just a side-effect of innovation.

This framing opens the door to future regulation under existing safety laws, where employers have a duty to identify and control psychosocial hazards. It also underscores her broader theme that “trust needs to be at the centre of policy” – workers must feel AI is being deployed with their wellbeing in mind, not simply as a tool to squeeze more output.

Five key principles for AI adoption

Across her recent speeches, Rishworth has sketched out five principles she believes should underpin AI adoption at work: trust, capability, transparency, safety and productivity.

Trust is foundational. Workers who see key decisions – about rostering, performance, or promotion – shifted from human supervisors to opaque software systems often “feel devalued”, she noted. That is corrosive not just for morale but for acceptance of AI more broadly.

Transparency and capability are the antidotes. Employees need visibility into how AI tools are used, what data underpins them, and what recourse they have if the system gets it wrong. At the same time, they must be given the skills to work effectively with AI, rather than left to muddle through.

Safety, for Rishworth, includes both traditional health and safety risks and the newer psychosocial harms linked to surveillance and intensification.

Productivity, finally, is the reason any of this is happening. Labor ministers – Treasurer Jim Chalmers and Assistant Minister for Productivity Andrew Leigh among them – have been vocal about AI’s potential to boost Australia’s sluggish productivity growth. But Rishworth is adamant that employees must share in the gains if AI is to be politically and socially sustainable.

Reflecting on past industrial revolutions, she noted that while living standards eventually improved, the initial waves of technological change were “a disaster for many” displaced workers. She warns that if Australians feel they are simply “training their own replacements” or being treated as “disposable”, AI will leave a “permanent injury” on the nation’s quality of life.

In practical terms, that means employers should strive to redeploy workers whose roles are automated, invest in upskilling, and ensure AI-driven productivity dividends show up in wages, conditions or reduced workloads – not just in profit margins.

Guardrails, not handbrakes

Rishworth has been explicit that she is “not going to close the door on regulation” of AI, particularly where psychological risks and fairness concerns are at stake. But she is equally insistent that any new rules must be carefully calibrated.

Her rejection of union veto powers and her defence of access to unfair dismissal processes indicate a preference for guardrails over handbrakes: setting boundaries for safe and ethical deployment without freezing the technology in place.

Enterprise bargaining, in her view, remains a “great mechanism” for giving employees agency over how AI is used, while still allowing employers the flexibility to innovate. With more workers now covered by enterprise agreements than ever before, there is scope for AI-specific clauses that address consultation, data use, training and job redesign on a sector‑by‑sector basis.

At the institutional level, the gap analysis of workplace laws, the AI Employment and Workplaces Forum, and ongoing consultation with the Fair Work Commission about managing AI‑generated claims all point to a phased regulatory pathway rather than a single, sweeping AI bill.

Humans as “master, not servant”

For all the nuance in the Albanese government’s position, Rishworth’s core philosophy is disarmingly simple: “humans must remain the master and not the servant” of AI.

That means ensuring workers have genuine input into how AI tools are adopted; that they are not reduced to mere cogs in algorithmically driven workflows; and that the technology ultimately serves human wellbeing and economic security.

Australia, she argued, is “in a good position to seize the opportunity from AI, rather than to crumble under its weight”. Early labour market data lends some support to that optimism. But union frustration over recent tech‑related job cuts and the rising anxiety of workers who fear they are “training their own replacements” are reminders of how quickly trust can evaporate.

The Albanese government’s emerging blueprint – tripartite dialogue, evidence‑based policy, guardrails on risk, and a firm insistence that AI’s gains be shared – is an attempt to shore up that trust before the next, more disruptive wave of automation arrives.

Whether it succeeds will depend less on the rhetoric at summits and more on what happens on the ground: in bargaining rooms, on construction sites, at the Fair Work Commission, and in thousands of workplaces where AI systems are quietly reshaping how Australians work.