Remarks to the Australian Strategic Policy Institute's Sydney Dialogue AI Masterclass

I want to say a few things about the Australian Government’s approach here on artificial intelligence.

But I am also very interested in participating in the discussion that’s been scheduled, so I don’t plan on giving you a long, discursive 45-minute address about the future of artificial intelligence. 

I understand you just had a very good discussion led here, but I do want to make a couple of points about where the Australian Government is on these questions and what the next few months should look like from our perspective. 

The first truism, of course, is that we live in a much more contested environment. 

We did when I was first elected to the Parliament in 2019. 

That is even more acutely so now, and the future of this technology is now a core part of what is going to shape that environment for Australia and for our partners. 

I think last year much of the focus of the discussion about artificial intelligence, of course, was on the economics of artificial intelligence, what it meant for investment and growth, what it meant for productivity in the Australian context, and while within government, we've been very focused on the security questions as well, much of the public debate I think really revolved around those propositions. 

Of course, artificial intelligence is not just about the economy; it's about our security as well, and it's about Australia's resilience in a much tougher global environment. 

And the Albanese Government's approach here, first articulated in the National Artificial Intelligence Plan that Andrew Charlton and I launched late last year, is capturing the benefits here in Australia in economic and strategic terms, making sure that we spread the benefits through the Australian community, so they are not just concentrated in the hearts of our capital cities and our financial districts.

But the use cases, [to use the] dreadful language that people use in this technology sector, are as available for people in Tamworth as they are in Turramurra. 

And, thirdly, that we build Australian capability and our approach to safety, and I saw that you had the director of the Artificial Intelligence Safety Institute with you just a few moments ago. 

So, we have announced a pathway to legislation. 

The Prime Minister's speech at the University of Sydney just a few months ago sets out a legislated approach on artificial intelligence standards and the Australian position in relation to copyright that is all about making sure that we capture Australian agency here. 

And, ultimately, the progress of that at this very important juncture in our history will be determined by the Parliament's capacity to support a legislative framework which delivers those outcomes for Australia on energy, on water, on the social dividend and economic dividend. Making sure that the approach that we take protects Australian copyright holders, but also with the objective firmly in mind, [that] there is a battle for technological supremacy across the world on artificial intelligence, and it's not this Albanese Government's approach that we should just be a cork bobbing on the ocean of that competition. 

It's our objective to secure Australian agency here and an Australian pathway through.

I think [about] the phrase “sovereign artificial intelligence”.

I think it's important to be clear about, in the Australian context, what that means. 

It doesn't mean every element of the technology stack, including frontier models, being developed here in Australia. 

That is not the approach. 

It is the approach to make sure that if you think about those imperatives that underpin our Future Made in Australia legislation, the same imperatives, what is our future competitive advantage and our capabilities? 

And, secondly, what is in the interest of Australian resilience and Australian security? 

Those two imperatives are driving an approach in artificial intelligence as much as they do in critical minerals and critical metals, and I do see this as just as importantly a security imperative as an economic imperative, and we are determined as a government to shape an Australian approach that gives Australia agency and Australia capability, not just from 2027 but 2037 on these questions as well. 

I'm very pleased that Parliament is agreeing to a joint select committee to examine some of these issues over the coming months. 

I want to see that be a vehicle for cooperation and constructive discussion across the Parliament. 

It's a very important national interest challenge for Australia. 

I might leave my comments there, Justin. 

Thank you for that welcome. 

I'm very sorry that because the Parliament's sitting up there, I would love to hang around for the day and listen to the discussion. 

I think Andrew Charlton said a little while ago this is about having the intellectual curiosity at a point where there is not the certainty around these technologies and technological progress that everybody wishes there was, because it's together that [we] will shape the Australian approach that delivers future Australian resilience and future Australian capability in this area.

Thank you very much.

Q&A

DAVID WROE, ASPI, HOST: Thank you so much Tim, that was great opening remarks. Let’s get straight into the big news story over the weekend about AI safety – how worried are you about some catastrophic risk arising from AI.

SENATOR THE HON TIM AYRES, MINISTER FOR INDUSTRY AND INNOVATION AND MINISTER FOR SCIENCE: I think firstly, you know the CEO of Anthropic, and which is supported by other technology CEOs and owners. What he said is pretty consistent, I think, with what he's been saying publicly and privately over the course of the last 12 months. I do take the safety challenges here seriously. The Artificial Intelligence Safety Institute is an important part of the Australian Government's approach here, building capability, monitoring safety in the Australian context, partnerships not just through the AI Safety Institute, but more broadly across the Australian system with our partners, both with the United States, but also with middle power economies like the United Kingdom and Canada, Japan, Korea. These are a really important part of our approach to safety. If you take a step back from this, these safety risks are absolutely there. The question for Australian policymakers and for Australians is how do we make sure that we build Australian resilience and capability in an uncertain world, and that's what the Prime Minister's speech was really directed towards.

HOST: I realise we're very close here. I feel like two kids on the back seat of the car. 

AYRES: We look like two galahs on a fence. 

HOST: Here we are. So, let's talk sovereignty. I'm interested in your vision for sovereignty success, say in 2030, 2040 kind of period. You've mentioned critical minerals, one strength that we can trade on, and this idea, I think, of using leverage, you know, with trade partners is a really reasonable one. Make ourselves indispensable. What other parts of the stack do you envisage us producing? What are you comfortable about relying on allies and partners for? And what should we actually worry about?

AYRES: Yeah, it's a good question, and the answer in 2034 or 2035 might be different as we advance and the technology evolves. But what I see in the Australian economy and the Australian technology sector at the moment – again, dreadful language, the technology stack – it catches on, right, and we all start using it. And I'm not sure that Australians really know what we mean.

HOST: I got asked what compute meant on TV recently.

AYRES: Yes, I had reason to look at Justin Trudeau's explanation of quantum as a part of inspiration for how we should approach these discussions. It is important, being really direct and clear with Australians about what these technologies mean for them. Yes, you're right to point to critical metals. In terms of frontier model development and training, location does matter, and securing training investment here in Australia is, in my view, crucial to shaping the future of the technology. But there are other areas of Australian capability that we have that I think most Australians wouldn't understand. 

I was at Syenta the other day that do chip packaging, advanced chip packaging. Again, explaining to Australians what that means, but an extraordinary set of technologies that they have developed, support from the Australian government there for them too. Our quantum capabilities are really important as these two technologies intersect. My objective here, the Albanese Government's objective here, is not to do everything, but to make Australia the indispensable country for our partners in the region and our partners more broadly. And that means that our approach to critical technologies – not just artificial intelligence, but critical technologies more broadly – is going to shape Australian sovereignty in years ahead.

HOST: Okay. One bit of compelling logic about indispensability is if we have frontier companies training their new AI models here, that gives us leverage to say we expect or even demand access to those frontier models at parity with US users and US organisations. Is that the way that you see it? And how do we actually make that happen? Are those commercial agreements or, you know, government-to-government agreements? How does that work? 

AYRES: Yes, it is.

HOST: And which one? 

AYRES: It'll be our all of the levers that we have at our disposal, from our FIRB approvals through to all of the other ways that we secure agreement. I've been impressed with how fast the major firms have been signing up to the Government's Expectations over the course of the first six months of this year. That does underscore the point here. Australia is a good place to invest. I don't need to bore this audience with the underpinnings of why we're such a good place to invest, but Australia should approach these issues, in my view, with some real confidence to shape the outcome here, not just to secure the investment, but to secure it in the Australian interest. 

Of course, for many of the issues that Australians care about now, those go to questions like electricity generation, water security, securing compute capability for our research system, but they must also go to being under the bonnet of these models, engaged in the security discussions and the capability discussions. And I'm very confident that the framework that we're developing gives us the tools to do that.

HOST: Speaking of ghastly terms, someone introduced me to the expression "radical optionality" recently, which is a very, very 2026 way of simply saying keep your options open but lean into all of them quite heavily so that you can jump on any of them. I feel that way about artificial intelligence because we know that it's huge, but we don't know exactly which way it's going. We don't know where it's going to end up. So, policy making is hard. I completely get that. Is it not then an argument for leaning hard and bold into, I guess, the obvious streams of development, whether that's hardware and semiconductors. I mean, models up to frontier models. Is there not a case for just being really bold and saying, you know, what we maybe we should turn that chip packaging into chip development? I know it's going to take 20 years, but we might look back in 20 years and say, geez, we're glad we did that. Or the same thing with frontier models.

AYRES: I don't want to evade that question. Firstly, just to say I had not heard the phrase “radical optionality” before. I'm going to give that some thought. But secondly, we do have in the critical technology areas that will intersect with artificial intelligence real capability in quantum, and you have seen the government leaning into the quantum capability that we have, which is world leading. And whether it's with the PsiQuantum investment in Queensland, with SQC [Silicon Quantum Computing], with Diraq, with a number of these kinds of quantum-related areas. That is an area where I see a straightforward path. Every one of these options won't be successful. There's a level of risk attached to all of this, but developing Australian capability here is right. We need to, I think, be open to the possibility that you've outlined, but not wedded to it, and I think that's consistent with the new radical optionality doctrine that you’ve now espoused.

HOST: I’ll add that to your annoying phrases book. I'll take the little win. The national security rationale, or the national security considerations, you've put some emphasis on that. I suppose, if we're talking about model development and, particularly, taking I suppose pre-trained models, fine-tuning them to our own unique economic needs, whether it's mining, agriculture, financial services, healthcare – the things that we are good at already as an economy – the best open-weight models right now are Chinese. From a national security consideration, what limitations should be put on Chinese or American models, for critical infrastructure, for the things that really, really matter?

AYRES: I can answer that question in two parts. We must have, do have, and will have an approach here that's shaped upon what is in Australia's interest and our security interests, and in the interests of our industry and our economy. And I'm not really prepared to go beyond that in this discussion. 

Secondly, I do think that what is not as well understood, while all the focus is on the large-scale models, Australia actually does have a competitive advantage in much narrower data sets. We've got a wealth of data. We've got real capabilities in mining and resources, health, agriculture, and the future economic advantage here for Australia can and should be in artificial intelligence embodied in technologies, in goods and services that Australia can develop here and offer the world, and our defence capabilities as well. 

And the strategic importance of these questions underscores how important it is that the second tier of the Australian Government's approach in the National Artificial Intelligence Plan is, that is to spread the benefits through the community. You would expect me to say that as a Labor minister, you know, with a social democratic approach to these questions, but securing the support of Australians in our regions and in our outer suburbs for this technology, and not falling for the trap that some other economies have fallen into, where it's you know there is some real push and pull, a lot of populism around these questions, is absolutely crucial. 

And what you've seen in Australia is a government that has taken the lead in advance of a number of these large investments to deal with questions like energy security and energy price in our electricity sector, and to make sure that artificial intelligence investment drives more generation development and transmission development and has an overall impact on the Australian economy and Australian competitiveness. That's of real benefit to everybody, and I think Australians want to see Australian governments acting in the national interest here, and that is why you know so much of our approach is focused on that.

HOST: Yeah, okay. Good. One last one from me before I turn to the audience. You've talked about economic harm and political disillusionment putting pressure on social harmony and trust in democratic governments. I've heard you talk about those things elsewhere. How worried are you right now about the risk that job disruption, whether it's permanent or temporary, let's say, you know, new value continues to create new forms of employment, and that hopefully is an endless upward cycle. Even if we accept that, there could still be five to ten years of pretty rough disruption. How worried are you that, especially if it is amplified by bad actors through disinformation and other tools that are now very, very easy to access through AI, that it ends in some form of bitterness?

AYRES: Well, I think I remember I was appointed as the Minister, and just a few days later I was at the AFR Artificial Intelligence Summit, and I made what I thought was an unremarkable observation that that trade unions should be at the top table and part of the discussion about artificial intelligence in the workplace, and in fact, the leaders of some major business organisations, in an unremarkable way, backed in this unremarkable assertion. We need Australians to be working together on these questions. The following day the headline was fantastic. It just said, “Tim Ayres is wrong”. I've never been so happy in all of my life. But I do think that that the way that we approach these issues, we have we have advantages in Australian institutions, and Australian culture, and our sense of Australian patriotism and national interest here that we should all be working together, and that is part of our Australian advantage. 

And we have worked together. I mean, I remember as a trade union official, while waves of robotics and automation were making their way through Australian manufacturing workplaces. We had a framework and institutional approach to making sure that workers were looked after in that process. Anybody who says with confidence that they know exactly what is going to happen in the labour market, you should just stop listening to me. There is some, I think, reliable work out there which indicates what is happening now in the Australian workforce and what is happening internationally, and it's remarkably consistent about the predominance of augmentation and changing workers' work rather than automation and robotics and replacement, but that doesn't mean that that is always going to be the case, and we should be alert to that. 

But I think giving our firms, our trade unions, our institutions, and our governments, state and federal, across our portfolios, the confidence and the capability and the frameworks within which to respond to these challenges is key and means we'll get a better outcome than economies that don't have that kind of mode of cooperation and contestability and challenge around these issues. I think Australians want to see not just in terms of their own jobs that they feel more secure about, but the jobs of their kids that there's an economic future for everybody.

HOST: Thanks, Tim. Who would like to ask a question? Yep, I can see Devon.

ATTENDEE: Thanks, David, and thanks, Minister, for an excellent conversation. The discussion in Australia around the Australian Standards for AI coming next year seems to be very focused on data centers. Meanwhile, other governments are legislating on AI companies themselves, whether it's risk management obligations, incident reporting obligations in the EU, California, New York, and elsewhere. Should Australia be joining with those partners in the Australian standards to address the risk from the companies rather than the data centers, and join these efforts towards collective AI governance? Thanks.

AYRES: Thank you, Devon. As I set out before, I am watching, and the Government’s watching, very carefully developments in those jurisdictions, particularly those large jurisdictions. And the AI Safety Institute working very closely with the safety institutes and other security institutions around the world. I'm really pleased with the progress that we've made and the cooperation that is happening there. 

We do think that there's a window for Australia where decisions about large frontier training model investment, data center investments will be made, and we want to make sure that we secure an Australian advantage here on Australia's terms. That is really what this legislation is about. It's quite carefully constructed the framework here that is making sure that it's secured on Australia's terms and making sure that we're delivering confidence from copyright holders about the way that this will work, and that it will be in no way a reduction in the copyright protections. So, that's where we're focused in terms of AI standards, but I think across government, our approach is about Australian agency and building capability as the core answer to this set of challenges. 

But we're watching very closely. We're very engaged with international partners, and as we've indicated in some of the social media questions, which aren't irrelevant to the artificial intelligence set of questions and the Digital Duty of Care, which is up on the Hill at the moment, we have been world-leading in terms of the approach that we have taken on technology regulation in the Australian interest, particularly when it comes to the interests of kids.

ATTENDEE: Hello, my name is Max Broad from the Lowy Institute. Bill Gates recently wrote an essay, and one of his recommendations was for a tax on tokens to be used to fund retraining of people whose jobs are made redundant and to fund the social safety net. Is the Government considering a tax on tokens?

AYRES: Where I start on these questions, I'm always slightly wary of an argument that large technology companies say that the consequence of the deployment of their technology will be large-scale job losses. Often, it's in their interest to make that argument, right? That we should resolve these issues by doing anything but securing additional investment and employment and growth in Australia. That is the Albanese Government's answer here: more jobs, more growth, more Australian capability in the interests of communities, but also in our strategic interest, and that's where we're focused at the moment. I do worry that there's a kind of loser narrative in that proposition that suits some interests and not others. We have to develop an Australian pathway here that's about making sure we're delivering an Australian economy that acts in the interests of Australians, not the other way around, and that we're shaping our future technological advantage in a pretty aggressive kind of way.

HOST: Right, Tim. We have arrived at our destination. It was great to chat with you.

AYRES: Thanks very much. 
ENDS.