On AI, Australia is finally asking the right questions

27 July 2026 | Stephen King

The Prime Minister's speech last week was a landmark in Australia's AI policy journey, both for its policy implications and because it showed the Australian Government is ready to lead a more complex conversation about AI in Australia.

Debates on AI are too often boiled down to 'for and against': pitting ‘tech utopians’ against a league of luddites. But the Prime Minister's speech highlights that AI cannot be treated as a single issue. To make sensible policy, we need to recognise that AI presents different challenges and opportunities at each level of the production chain – from data centres, to model training to AI deployment – and in each area of the economy.

When it comes to the use of AI, the debate is not about whether we should regulate but how we should regulate. The government is systematically reviewing our laws, checking that they cover AI harms and, where they don’t, filling the gaps. This ‘review and adapt’ policy builds on the strength of our existing laws and regulators, making sure they are fit for the AI age.

The Productivity Commission strongly supports this approach. It recognises that the potential for bad actors to use AI is everywhere, and that every regulator and every law needs to be ready. Trying to cover all the harms and benefits of AI in a single set of laws is naïve and risks creating complexity that will stifle our AI adoption.

But the use of AI is only one part of the stack.

Data centres are a critical part of the AI supply chain, providing the ‘computer grunt’ needed to both train and run AI models and applications. As the ‘National AI Plan’ notes, Australia is one of the favoured destinations worldwide for investment in data centres. But such investment needs to be in the interest of the broader community.

The government is moving towards a national approach to data centre investment. This makes sense. Too often, differences in state and territory rules stymie investment in both physical and human capital. The Productivity Commission has long called for consistent national approaches in other areas such as skills recognition and transport.

A national approach will avoid policies where one state attempts to seize the benefits – or avoid the costs – of data centre investment through regulations that harm Australia’s overall interests.

The government has also recognised that the cost of building new electricity or water infrastructure to support data centres should not be passed on to the broader public. And the government’s moves to increase the clarity and speed of approvals processes is something the Productivity Commission has called for across the economy. These new legislated national standards for data centres should be targeted and proportionate: they cannot be expected to solve every policy challenge, or they risk becoming so burdensome that it deters the investment Australia needs.

Then there is the training of AI models. Here data is a key input. The government’s move to a ‘tell us once’ approach for government data is a critical step towards best practice. The Productivity Commission has also argued that Australia’s broken privacy laws need an overhaul to ensure those who have our identifiable data have a duty of care over that data.

In our report to government in December last year, the Productivity Commission recommended that the government should not rush to change our copyright laws. The current laws underpin licensing agreements that properly reward Australian creators. They are not perfect and licensing may not cover all the material needed for AI training. But AI-related copyright exceptions, which are currently the subject of a huge number of legal challenges in the countries that have them, are unlikely to give tech firms the clarity they seek. In this area, it is important that Australia takes the time to learn from other countries rather than rushing to solutions.

Establishing an Office of AI at the heart of government is common sense. AI policy is broader than any specialised department and will cross department boundaries. For example, recent moves by the US Government to limit access to frontier AI models highlighted trade, security, and resilience issues for Australia.

However, supply security is not a new issue. And we are not alone. Australia should work with like-minded ‘middle powers’ to ensure a secure AI supply chain. We do not need an expensive, duplicative Aussie-AI model. But we do need international agreements and protections around the standards and availability of AI models, led by Australia and our close allies.

So where is Australia’s AI opportunity? Our strength in software-as-a-service points to one possibility. We build and export key business and consumer applications worldwide and these same skills will be needed for AI applications – what some are calling Model-as-a-Service.

But here, as with so much else in AI, we can only make an educated guess. What’s certain is that to capture the benefits the government needs to take the right policy approach: protecting us against the harms as they exist, while holding firm against calls for knee-jerk regulation or nice-to-have conditions on investment that see us fall behind. The Prime Minister’s speech hasn’t given us all the answers, but it shows that the government is asking the right questions.