Australia’s new AI safety priorities seen as fragmented and underfunded

Australia is a useful bellwether for how mid sized democracies are struggling to turn big AI rhetoric into coherent policy. This commentary makes the case that Canberra has moved from a serious, if slow, attempt at targeted AI regulation toward a looser, more aspirational agenda built around an underfunded AI Safety Institute and a not yet delivered digital duty of care. The through line is fragmentation: privacy reform over here, automated decision making safeguards over there, consumer law analysis somewhere else, all without a strong centre of gravity or enough money to test and audit frontier systems properly.

From an AGI race perspective, that might look like a local story, but it echoes a wider pattern. Many governments want the economic upside of AI, worry about being seen as ‘anti innovation’, and are therefore defaulting to voluntary frameworks and rebranding old initiatives as “AI safety” rather than grappling with enforcing hard constraints on powerful models. That puts more weight back onto the internal governance of a handful of labs and hyperscalers. For companies choosing where to deploy frontier systems at scale, Australia’s approach may feel permissive in the short term, but the lack of clear, well resourced rules could create more political risk if a high profile incident forces a sudden clampdown.