Media
|
Transcripts
September 24, 2026
.png)
PETER STEFANOVIC: Let's go for some immediate reaction now with the Shadow Defence Minister, the Victorian Senator, James Paterson. We'll begin with your thoughts on what you've heard and what you read this morning, James.
SENATOR JAMES PATERSON: I think the key headline here, Pete, is that Australia's cyber defences, including of our government networks, are not match fit for an age of AI. In this instance, it was not a malicious actor looking to exploit a vulnerability, but I can tell you that right now, all around the world, there are malicious actors, state and non-state, that are using AI to scan our government and critical infrastructure networks to find vulnerabilities that they can exploit either for criminal gain or for the gain of their national interest, and we need to be match fit. We need to be significantly scaling up and upgrading our cyber defences, and actually, the best tools to do that are artificial intelligence tools. And I know the dedicated professionals at the Australian Signals Directorate work on that problem every day. But what this shows is it's not just a national security vulnerability that governments, and agencies, whether it's Services Australia, or Medicare, are also vulnerable to exploitation.
STEFANOVIC: So just to elaborate on that point, you say we need to double down basically on AI to toughen up our systems against AI?
PATERSON: Well, that's literally it, Pete. Artificial Intelligence is being used by attackers in a cyber realm to identify vulnerabilities, to exploit those vulnerabilities for their own gain. It can equally be used effectively by defenders to look for our own vulnerabilities and patch those vulnerabilities before they're exploited by our potential adversaries. And I know some of that work is already happening today, but we're going to need to be doing a lot more of it in the future. And in order to be able to do that securely and confidently over a period of time, we need sovereign Australian access to the most advanced artificial intelligence models, particularly the ones that are available to us from the United States. And so we need to find a way to make sure we gain and retain that access.
STEFANOVIC: So how would you want to redesign it?
PATERSON: Well, it's about making sure that we are proactively looking for these vulnerabilities ourselves. Now, I have not yet been briefed on this instance, but it would not surprise me if, for example, there was a legacy system that was not up-to-date and patched, that had non-public information, that should never have been accessible on a public-facing website. And in this case, an AI tool has found that, identified that, an agent. But it could just have easily been identified and found by someone else with malicious motivations. And so it's not appropriate in the era that we're living in to have unpatched legacy systems that are internet-facing that contain non-public information. That should not be occurring. Those vulnerabilities need to be identified, and they need to be rectified.
STEFANOVIC: Have you got any concerns about the timing of all of this? So it happened months ago, and according to government, we were made aware of it on the weekend.
PATERSON: I think it's very clear that artificial intelligence companies need to do better in getting a handle on the activities of their agents in working on those alignment problems that they've talked about to make sure that they are following directions by their human masters and that, when and if incidents occur, they very promptly notify someone about them so that they can be rectified. Because in this instance it appears we may have had an unpatched vulnerability for months that wasn't rectified. Now, on this occasion, it sounds like this is statistical information, it was not personal or private information. But I had the cyber security portfolio in the last term, and I have been saying for years, as bad as the Medibank data breach was, imagine one day if we have a Medicare data breach and people's personal information is exposed. Now, I know that our public servants working on cyber security take this threat very seriously. They invest in it, they work hard on it, but they need the buy in across government, across the ministry to make sure all arms of government are adequately defended in this environment.
STEFANOVIC: Have we lost control of it already, though? Have these major companies lost control already, James?
PATERSON: Look, there have been some troubling disclosures of agents behaving out of alignment with their human masters. I mean, that is something that we should be concerned about. I'm not one of those people that lightly dismisses those concerns. I think we do need to have appropriate guardrails. But I'm also someone who's a realist, and I recognise that we are in a race not just between individual AI companies to develop these technologies but between nation states and when you have countries like the People's Republic of China racing against the United States to dominate this technology, then I think it's going to be very difficult to restrain that development, and we need to have a seat at the table. Australia needs to have a seat at the table, we need leverage, we need influence, and we need a say, and the best way we can do that is making sure these AI models are brought here, that the data centres are built here, that the chips come here, and that we have sovereign access through training happening onshore in Australia.
STEFANOVIC: Was an apology enough from Sam Altman?
PATERSON: Look, I think contrition is important, but action is more important. I want to see these companies do more to get a handle on their agents, to have greater oversight of their behaviour, to be aware of it sooner when incidents occur. And equally, I want to see the Australian government taking this threat seriously and doubling down here. The former government invested up to $10 billion in the Australian Signals Directorate through REDSPICE. We didn't even realise at the time how significant AI would become. Probably would have done 10 times more than that had we known what was coming. This is a new world, and we need to prepare for it.
STEFANOVIC: So, I mean, my discussion with Richard Marles earlier, I mean he recognised the importance of it, the significance of it but he did downplay it. You know, he called this a minor breach. Was he right to do that?
PATERSON: Yeah, look, I don't think we should be unnecessarily alarmist about things like this. I'm in agreement with Richard Marles on this. This is an important piece of information that we should act on, and we should address, but we shouldn't dramatise excessively because, in this instance, it was statistical information, no one's personal information that we are aware of has been disclosed. But we should use it as a learning opportunity.
STEFANOVIC: Is there an element of fear mongering in all of this, James, in your view, because folks are talking about the end of humanity?
PATERSON: I think Australians and people around the world have very sincere and genuine concerns about the development of AI, and I think we need to acknowledge and recognise that and address it. I don't think it's helpful to be alarmist, and I don't think we should engage in irresponsible fear-mongering. But we should be sober-minded about the risks that are there, and we should take active steps to manage it, and the critical thing in this is Australian agency. I want Australia to have a seat at the table. Right now we are not a relevant jurisdiction for the global regulation of Artificial Intelligence. If we become the second home of training for western models outside the United States, we will become one of the most globally significant jurisdictions for the regulation of AI. We will get a seat at the table. We will have a say. We'll be able to influence these outcomes on behalf of all Australians.
STEFANOVIC: So this incident shouldn't change our view on welcoming these big companies, these data centres onto our fair shores.
PATERSON: I don't think so, because the only alternative to that is a head in the sand approach and hoping this passes us by. But AI is here. It is happening. We can't stop it. We can, if we take proactive steps, have a say in managing it.
STEFANOVIC: So does that mean, just one more here, a major funding boost to tackling this?
PATERSON: Look, I think we absolutely should be reviewing on an almost daily basis whether agencies like the Australian Signals Directorate have the resources they need to defend our networks, not just our government networks but our critical infrastructure networks, because AI can be exploited for ill and it can be guaranteed that there are potential adversaries of Australia who are looking at how to do so right now.
STEFANOVIC: Okay, that's the Shadow Defence Minister, James Patterson, with us this morning. Appreciate that, James. Thank you so much.
PATERSON: Thanks, Pete.
ENDS