Edited by 
Short answer
Australia's main body for advanced AI risks is the Australian AI Safety Institute (AISI), announced in November 2025 with $29.9 million in funding and operating since early 2026 inside the Department of Industry, Science and Resources. It tests frontier AI models before and after release, studies emerging risks and advises government, but it does not regulate. Under the December 2025 National AI Plan, Australia chose not to pass a standalone AI Act, relying instead on existing laws such as privacy, consumer and online safety law. Australia has no specific policy on superintelligence, but the institute's frontier model testing is the closest thing to one.
Why AI safety matters to Australians
Australians use the same frontier AI systems as the rest of the world, built mostly by US companies, so decisions made in Silicon Valley shape what reaches Australian schools, hospitals and workplaces. Surveys have consistently found Australians among the more cautious publics on AI: in the University of Melbourne and KPMG global trust study, Australia ranked near the bottom of surveyed countries for trust in AI and acceptance of it, with most Australians saying more regulation is needed.
That caution is one reason the government created a dedicated institute rather than leaving AI risks to existing regulators alone.
The Australian AI Safety Institute
| Detail | What we know |
|---|---|
| Announced | November 2025 |
| Funding | $29.9 million for establishment, early operations and testing infrastructure |
| Started | Early 2026 |
| Home | Department of Industry, Science and Resources |
| Leadership | Reported to be led by Dr Kate Conroy, with Prof Paul Salmon leading safety science research |
| Role | Advisory, not a regulator |
| International | Member of the international network of AI safety and measurement institutes, alongside the UK, US, Canada, Japan, Singapore and others |
What the institute does
- Tests frontier models: voluntary testing of the most capable AI systems, before and after they are released, with technical partners. In July 2026 the government said testing had begun, with projects alongside the Gradient Institute on AI agent risks and CSIRO on alignment.
- Tracks emerging risks: from everyday harms like scams, deepfakes and biased decisions to higher-impact frontier risks such as cyber attacks and loss of control.
- Finds gaps in the law: identifying where existing Australian laws do not cover new AI risks.
- Shares information: working with regulators such as the eSafety Commissioner and the OAIC, and with partner institutes overseas.
The National AI Plan
The National AI Plan, released in December 2025, sets Australia's overall direction. It has three broad goals: capturing the economic opportunity of AI, spreading the benefits across the community, and keeping Australians safe. The AI Safety Institute sits under the safety goal, while the National AI Centre helps businesses adopt AI.
The plan's most debated decision was to drop earlier proposals for mandatory "guardrails" on high-risk AI in favour of using existing, technology-neutral laws, updated where needed. Supporters say this avoids slowing innovation; critics say it leaves gaps for powerful general-purpose AI.
Does Australia have a policy on superintelligence?
Not explicitly. No Australian law or plan mentions superintelligence by name, and the government has not taken a position on proposals such as the international Statement on Superintelligence, which calls for a prohibition until superintelligence can be built safely. Some Australian academics are among its signatories.
In practice, the AISI's work on frontier model testing and loss-of-control risks is where Australian policy meets superintelligence. Because Australia does not build the largest models itself, its influence comes mainly from testing, international cooperation and the conditions it sets on AI used in the Australian market. Read what is superintelligence and AGI vs ASI for background.
How Australia compares
| Country or region | Approach to frontier AI |
|---|---|
| Australia | AI Safety Institute (advisory); no standalone AI Act; existing laws |
| United States | Center for AI Standards and Innovation tests frontier models; federal focus on AI leadership; state laws in California and New York |
| United Kingdom | AI Security Institute, one of the largest government AI testing teams |
| European Union | EU AI Act with binding rules for general-purpose AI models, including the most powerful |
| Canada | Canadian AI Safety Institute; no federal AI act in force |
Read should superintelligence be banned? for the US debate in more detail.
AI hardware and data centers in Australia
Australia is also becoming a location for AI compute. Global cloud providers and Australian operators are building large data centers, especially around Sydney and Melbourne, and the National AI Plan supports more local AI infrastructure. That raises local questions about electricity, water and grid capacity. Read what is an AI data center for how these facilities work.
Frequently asked questions
Does Australia have an AI Act?
No. Unlike the EU, Australia has not passed a standalone AI law. The December 2025 National AI Plan relies on existing laws, such as the Privacy Act, Australian Consumer Law and the Online Safety Act, plus the work of the AI Safety Institute.
Is the Australian AI Safety Institute a regulator?
No. It tests models, researches risks and advises government. Enforcement stays with existing regulators such as the ACCC, OAIC and eSafety Commissioner.
Who leads Australia's AI Safety Institute?
Reports in 2026 name Dr Kate Conroy as its head, with Professor Paul Salmon leading safety science research.
Can Australian researchers work with the institute?
Yes. The institute partners with research organisations such as CSIRO and the Gradient Institute and has recruited technical staff from the Australian AI research community.
Sources and further reading
- Department of Industry, Science and Resources, National AI Plan and AI Safety Institute
- Herbert Smith Freehills Kramer, Australia to establish AI Safety Institute (November 2025)
- Let's Data Science, Australia tests frontier AI models for safety (July 2026)
- University of Melbourne and KPMG, Trust, attitudes and use of artificial intelligence: a global study (2025)
Leave a Reply