Near-term AI policy priorities for the Federal Government

I'm an independent Member of Parliament in Federal Parliament from Western Australia. And in May I put out a discussion paper called Shaping the Future of AI for Australia, which covers 18 policies or issues that I think the government should be acting on right now. And I'd really love to get your feedback on this. So the thing I'm hoping to get out of today is a bunch of people who have great knowledge in this area, [inaudible] giving me some feedback into whether we're on the right track and what the priorities are.

As an independent, I don't have to toe a party line. I'm not in government. I'm not in either of the major parties. What it means is what I can do is put issues on the agenda and try and progress the issues, especially if they're a bit hard or we don't quite know what to do with them. I don't have to do what my party has always done before and remain consistent to a previous approach. I can just go out and look at the evidence and say, what is the issue and what does the government need to be doing right now?

And AI was one of the issues where I thought this was an area where it was being missed, and we really were not acting at the speed that we need to be acting. And I'm not going to go into all the impact of what it could do — I think we're all well across that. But I think when it comes to government and how government approaches it, it's a little bit like climate change in that it's sort of too big to think about, and therefore it's hard to know where to start.

So the question that we went into this with was, what's the role of government right now in managing this transition? And I engaged with my community in Western Australia, which is probably like many other communities in terms of what issues came up, [inaudible] and then also spoke to about 40 key stakeholders, many of whom are here of the there's couple of days to pull together this paper. And the key message really is that government needs to take a more active approach. Now, I'm pleased to see the AFR reporting this morning that the government's decided to take a more active approach, which I, of course, take full credit for.

The other message that I heard from community was also this fear that our future is being shaped by a bunch of billionaires who live overseas rather than our values and our voices. So I want to run through the paper within the limitation of my ten minutes very quickly, and then try and also leave some time at the end for questions. But firstly, what I found in my community: people were four times more likely to be concerned than optimistic about AI. Eight out of ten didn't think government was doing enough.

Eight out of ten didn't think that current policy would ensure that AI benefits Australians rather than overseas tech companies. So this is really a community concern that government needs to act on. Now, of course, [inaudible] uncertainty makes policy development really hard, and a lot of conversations about AI governance end up being quite abstract and theoretical. So we tried to really force the conversation down to something concrete — a list of near-term priorities that are worth doing regardless of how the frontier develops. So I've got 18 policies organised into five categories: getting the structures right, realising the opportunities, dealing with the short-term harms, dealing with the longer-term emerging risks, and making sure that we share the benefits of AI.

And it is a discussion paper, not a finished platform, and I genuinely would love to have your input. The first part of that is setting up the structures. And with AI policy, we need institutions that can actually move at AI speed, which government generally doesn't. There are good reasons that government moves slowly when it comes to legislating, but as the world changes, we're actually going to need more nimble structures to address that changing world. So the three policies here are significantly increasing funding for AISI. And we've heard some about that this morning.

We have the potential to be really leading on that. Also to do the same for the National AI Centre, which is more focused on the opportunities. And then stress test our existing laws, whether it's from child sex abuse material, critical infrastructure, legal liability. There are lots of AI gaps in our current legal framework, and we really need to do that work fast. Focusing on the first one, the Safety Institute — we heard a bit about the role that it plays this morning. Right now, it's funded at under $8 million a year, which might sound like a lot of money, but it's really not.

The UK's equivalent, the AI Safety Institute, gets 15 times that. Now, money isn't everything. But we are proposing a $100 million investment in the AI Safety Institute, and if we did that, we could have an outsized impact when it comes to our role in shaping the future. So it seems like a good place to invest. The next are about capturing the opportunities, and we can't assume that the benefits will materialise. Many stakeholders are telling me that the government should be playing a more active role in identifying the opportunities and creating the conditions for them to be realised.

So the four policies here: establish a National AI Missions program, addressing the copyright problem, establishing an AI for Science program, and investing in AI skills and talent. I want to speak a little bit about the AI missions program. We have to be realistic about what we can do here. It's not going to be everything. And being clear-eyed and coordinated about where our strengths are will mean we're able to unlock those benefits in a time frame that's realistic. So my proposal would be that the National AI Centre runs a structured process to select a handful of what we're calling missions — areas where we really coordinate our efforts.

And that would be testing against whether there's an advantage there already, if something is nationally important and if it's technologically viable. So, for example, we're likely to have some expertise in natural disaster prediction, agricultural resilience, health — using AI with Medicare and PBS data — clean energy transition, possibly data centres. To really work out what are the things that we can do well and line up the ducks so that we're going full tilt where we have the potential, and concentrate our limited capability behind a small number of areas.

The next are around dealing with current harms. And this is the area that is of most concern to community right now, and what I hear about when I'm door knocking. People see these risks and they worry about what harm is being done. Now, digital duty of care is my number one priority in this area. Extending the under-16 social media ban to cover... I'm not going to read them all. But digital duty of care — I think this is a really important one because it's a philosophical shift.

When we walk into a shop, we expect it to be a safe place, but we don't have the same attitude to [inaudible] online spaces. And if we can get that right for social media, which is present in people's minds, it starts that cultural shift to thinking differently about the responsibilities of platforms. Emerging risks: all the stuff that we hear about that was scary to hear — that there's a more than 10% chance of all those bad things happening. Scenario planning seems to make the most sense here, so that you're not asking the government to say this is the disastrous direction we're going into, but at least to say we're onto it.

We've thought about what can be done. Establish a mandatory transparency and incident reporting scheme, and also build AI into our crisis management framework. And lastly, sharing the benefits. So national AI literacy programs, binding obligations around data centres and working out how to actually tax international AI companies. And this is, I think, where we've seen this movie before with offshore gas, where we lock in generous terms early to compete for that investment and then fight for years to claw back a fair share. So we've got to make sure we're thinking about that from the beginning.

So I think it's really important that the government starts acting with urgency on both the promise and the peril. And I think these are no-regrets steps. So this is what I'm working on, trying to get the government to focus on. And a little bit of advice just to end on, for those in the room who are doing advocacy and are engaging with policymakers — and I suspect it's a lot of people here — please try and be as practical and specific as you can in making recommendations, because for policymakers, they're not experts.

And being presented with a problem is much harder than being presented with some suggested solutions. Even if it doesn't end up being the final solution, it gives you something to work with. So we need all the brains feeding into the policy development, and we'd need it to be as practical as possible. Please go follow the QR code, [inaudible] have a look at it and let me know what I've missed. Let me know what deserves more emphasis. And if you disagree with a direction that we've proposed, please leave that feedback.

You can also come and have a chat to me over the next few days. Or my policy advisor, Heath Hasemer, who's over here. Please find one of us. Give us a wave, Heath. Gone. Come and find one of us and tell us what you think. Really keen to get some feedback so we can keep pushing government to act on those 18 areas. And there are some copies — hard copies, for those who still do hard copies — at the front of the room. Please come and grab them.

Thank you.

Greg Sadler

Thank you. Kate, we might try and squeeze in one question. I know that there's a lot in the chat, so maybe the top question was: you're one of the few MPs who sort of had this genuine interest in this AI safety policy question, which has obviously motivated us all to be in this room today. So what motivated you to pick up this as an issue that's sort of central to the things that you're talking about?

Kate Chaney

I think at the beginning of this term, when I was thinking, what am I going to focus on in the next three years? This was a topic that I thought, it's not top of the list for community concerns at the moment, but I think by the end of the term it will be. And there's always a balance between do you just respond to what you're hearing, or do you try and pre-empt what you think you'll be hearing? And so far, I think it's right. One of the really hard parts is doing policy work in an area where it's all novel.

It's not like you can just roll something out that's worked overseas. And so I think it's much easier to do policy where there's something very specific. But saying here are 18 things — it's high level and hard to work on. I think there is interest. But without those practical suggestions coming from civil society and other sources, it's hard for politicians to grasp onto it, especially when it's so easy to look like an idiot in this area. So assume low knowledge if you're talking to politicians about it.

And try and be as practical as possible.

Greg Sadler

Great. Thanks very much.