Australians worry more about AI misinformation than job losses, survey finds

Two-thirds of Australians want the country to move more cautiously on artificial intelligence, but their biggest worry is not job losses or data centres. It is misinformation and deepfakes, according to new polling shared with POLITICO.

For Australian business leaders building or buying AI systems, that distinction matters. Public anxiety is not aimed at automation itself. It is aimed at trust, transparency and who controls what AI-generated content people see. That has direct implications for how AI and enterprise resource planning (ERP) tools get positioned, procured and governed inside Australian organisations.

According to POLITICO, JWS Research surveyed 1,230 Australians between 6 and 10 August 2026. The results show 44 percent view AI mainly as a risk, against 17 percent who see it as an opportunity. Another 28 percent are conflicted, and 10 percent are unsure.

Tom Cameron, JWS strategy and insights manager, told POLITICO the public is more cautious than fearful, with roughly half wanting a slower rollout and only about a fifth pushing for speed.

The breakdown by group is notable. Women over 55 were the most wary, at 56 percent. Men aged 18 to 34 were the most optimistic, at 31 percent. Greens and One Nation voters, despite sitting at opposite ends of the political spectrum, were closely aligned on caution.

The standout finding is what worries people most. Cameron said the top concern, and the top issue people want regulated, is misinformation and deepfakes, with privacy and data misuse close behind. Job losses ranked lower than both.

Almost a third of respondents, 29 percent, also worried about AI making harmful decisions without human oversight. According to POLITICO, that concern is shaped by memory of the Robodebt scheme, an automated welfare debt-recovery system that wrongly targeted hundreds of thousands of Australians. Robodebt did not use AI, but it left a lasting mark on public trust in automated government decisions.

Businesses weighing where to draw the line between automated and human decision-making can work through this systematically. A structured AI Feasibility Analysis helps identify which processes genuinely need human sign-off before deployment, rather than defaulting to full automation and discovering the gap later.

Unlike in the United States, community backlash against data centres has not taken hold in Australia yet. Fewer than 1 percent of respondents named data centres unprompted when asked what government should focus on, and only 3 percent said they had no concerns at all about them when asked directly.

Where concerns do exist, they are practical: electricity demand and prices first, then water use, then environmental impact. Cameron noted that around a fifth of respondents had never heard of data centres at all, which he called a first-mover opportunity for whoever defines the issue first, government, industry or campaigners.

The polling lands as the federal government moves on several AI fronts. Prime Minister Anthony Albanese announced in July that large data centres will face new standards, including obligations to underwrite their own power supply and cover connection costs. A new Office for AI has also been established, alongside a joint parliamentary committee examining AI's effect on national security, productivity and living standards.

For CEOs, CTOs and founders deploying AI or ERP systems in Australia, this data suggests customer and staff resistance is less about AI replacing people and more about content integrity and decision transparency. Products that can demonstrate clear provenance, audit trails and human oversight on automated decisions are likely to face less friction with staff, boards and the public than those pitched purely on efficiency gains.

The Robodebt-shaped caution around unsupervised decision-making is also a governance signal. Any AI or ERP deployment touching customer records, payments or eligibility decisions should build in visible human checkpoints, not just as a compliance box, but as a trust feature worth mentioning in vendor conversations and board reporting. Building this in from the start, rather than retrofitting it, is typically the difference between a smooth custom AI development project and one that stalls at internal sign-off.

The joint parliamentary committee established in August will examine AI's impact on national security, productivity, competitiveness and living standards, though no reporting date has been confirmed. Separately, former Electoral Commissioner Tom Rogers told POLITICO that Australia's next federal election, expected between late 2027 and early 2028, will face an unprecedented level of AI-driven mis- and disinformation, adding urgency to the regulatory debate ahead of the data centre standards taking shape.