A leading AI researcher recently warned that Australia’s retreat from creating a formal advisory body of experts risks leaving young people exposed to harms driven by commercial incentives. Instead of that advisory panel, the government issued a National AI Plan that prioritises investment in data centres, telecommunications, and workforce training, proposes an AI Safety Institute (now recruiting), and offers limited transparency measures for public sector AI. So far, disclosure has been patchy.
How does this fit into the wider picture of AI regulation? International responses vary. The EU’s AI Act restricts certain harmful uses, for example those that exploit vulnerable groups, but is struggling to operationalise rules for many high‑risk systems. In East Asia, South Korea, Japan and Taiwan have enacted new AI laws that give authorities stronger intervention powers, and industry resistance is expected. Other big players, notably the US and UK, lack comprehensive, all‑embracing AI statutes: the US focuses on controls for federal use while curbing state-level regulation of private actors, and the UK has leaned on technical standards and a new safety/security agency rather than sweeping legal reform.
These differences reflect a common policy tension: when risks seem remote regulators are reluctant to act, but once harms become vivid, regulatory change is costly and slow. Australia also lacks the international leverage enjoyed in other sectors, which reduces its influence on global norms. Given these uncertainties, predictable regulation is important for businesses and researchers. Australia’s Assistant Minister for Science, Technology and the Digital Economy has acknowledged the need for clear principles and wide buy‑in as an insurance policy against future harms.
The National AI Plan frames safety primarily by asserting that existing legal frameworks can be applied to AI — for instance, consumer protection laws prohibiting misleading conduct could cover harmful AI outputs. That view aligns with earlier government statements and some official reviews, but many legal and technical experts argue it is insufficient on its own.
There are concrete reasons for concern. AI systems can be complex, partially autonomous and opaque, which makes it difficult to assign liability or responsibility under traditional legal tests. Those gaps were flagged as early as 2023, yet there has been no systematic program to close them. The regulatory landscape inside Australia is also fragmented: researchers have counted at least 21 mandatory or quasi‑mandatory policies governing government use of AI across federal and state levels. Courts have had few opportunities to settle key issues — there are almost no precedent‑forming cases in negligence, administrative law, discrimination or consumer law involving AI.
The plan promises ongoing monitoring and intervention as challenges arise, but it is vague on how monitoring will be funded, who will coordinate it, and how agencies will be empowered in practice. Saying that every existing agency will take responsibility for AI is not the same as giving those agencies the resources, expertise and statutory authority needed to manage privacy, consumer harm and discrimination risks.
Looking ahead, political and regulatory shifts overseas could alter pressure on Australia. A largely hands‑off stance does not automatically produce regulatory clarity when many processes are delayed or incomplete. The government seems to expect courts, regulators, businesses and citizens to retrofit older laws and institutions onto a rapidly changing technology landscape. Aside from a possible boost to public‑sector automated decision‑making rules after prior inquiries, the overall posture is effectively wait and see. Whether that approach will protect people and provide sufficient certainty for innovation remains an open question.
