Media
|
Transcripts
September 28, 2026
MELISSA CLARKE: We have the Shadow Minister for Defence, James Paterson, with us this morning. James, welcome to Radio National Breakfast.
SENATOR JAMES PATERSON: Good morning, Mel.
CLARKE: Now we've seen OpenAI confirm breaches to government websites here in Australia and now in the US as well. We've seen that several US government departments have been inadvertently hacked too. How extensive do you think this problem is?
PATERSON: Look, I'm not at all surprised that we are learning more about this because we know that the AI companies have what they call an alignment problem. That is, their AI models are doing things that they were not explicitly tasked to do by their human masters. And we know from the Hugging Face incident that this has the potential to be quite serious. I think it's important, though, when we talk about this that we're really clear about what has happened, what hasn't happened, and we shouldn't be engaging in unnecessary fear mongering or alarmism because that risks a real boy who cried wolf situation. Let's talk about the serious threats when they're serious and the more modest ones when they are more modest.
CLARKE: Now, I guess the question here is the desire to get ahead of potential threats and given we are finding and discovering these smaller breaches, it does raise the question that there's probably more that we don't know about. Are you confident that the tech giants like OpenAI and Anthropic are being upfront about everything that they do know, given we've seen delays in getting that information to governments?
PATERSON: I think it's very clear that they are trying to get on top of this issue, that they're grappling with this very fast evolving technology, and they may themselves not have full visibility of what these models have done and that is worrying, and we should treat these incidents in Australia as a free warning because the next time it could be much more serious. I think we should be having a conversation about mandatory notifications because in this instance, OpenAI voluntarily disclosed it to the Australian government. It's possible there have been other incidents that the government has not detected itself and have not been voluntarily disclosed. And there are people with malign intent looking at our government systems, looking at our financial systems, our electricity networks and thinking about how they can use this technology to exploit it. And I guarantee you they will not be telling us about it.
CLARKE: The government's rapid review is also looking at the question of legal liability. We heard from Katy Gallagher just then that there is a need to look at how to deal with these questions around responsibility for new technology, but many in the Coalition are raising concerns that this could amount to threatening AI companies and could impinge on cooperation with them. Do you really think that's a problem? Surely these AI giants understand that as new technology comes about, governments have to consider the legal framework around them?
PATERSON: Liability is a really important and difficult question to determine because our criminal law relies on intent, and an agent doesn't have intent. And if it's misaligned with its human master, it might not even reflect the intent of the person who programmed it. So this is a thorny and important policy issue to resolve. But I have to say, some of the conversation about criminal penalties, about the most recent incident, are too silly for words. I mean, this is data which is prepared solely for the purpose of being publicly released. It is not even clear that any actual hacking was involved in accessing it. There are credible technical experts out there saying that this was freely accessible to anyone who knew where to look. If a teenager replicated this exact data access, there's no way that we would charge them with anything, and they could do it happily from their mum and dad's basement on a basic laptop without an AI agent. So let's talk about criminal liability for a really serious incident, not for a trivial one.
CLARKE: Are there questions around negligence and the role that companies have in their responsibilities that need to be addressed too?
PATERSON: I think so, and also independent assurance and evaluation. I think Australians and anyone in the world is entitled to demand an independent oversight of these companies because they are powerful, because their technology is evolving very quickly and because they themselves say they don't have a complete handle on what these models are doing. So Daro Amodei talked about external independent embedded evaluators. I think that's a good idea. I think we should look at how we require that by attracting AI companies here in Australia.
CLARKE: So you think that should be a condition of having third party independent evaluators for us to support development of data centres, for example, here in Australia? You want to make that a condition?
PATERSON: I think we have an opportunity to strike a really good deal here for Australia that gives us strong oversight and evaluation, that gives us independent assurance, that gives us some confidence about safety and also attracts billions of dollars of investment, gives Australia access to the world's most advanced models and allows us to use it to defend our networks from the attacks that we know are coming in the future from AI agents. So there is a package deal to be done here, and I think it should be seized.
CLARKE: You're listening to Radio National Breakfast where the Shadow Minister for Defence, James Paterson, is my guest. Just briefly on a few other issues, we've seen unfolding overnight our time that authorities in the UK have thwarted what appears to be a terrorism plot at an air force base, the RAF Fairford in the Cotswolds. Have you had any briefings about it, noting this is at very early stages?
PATERSON: Not yet, but I'm troubled by the media reports that I've seen, which do look very serious. Let's wait and see for official statements from the UK government, but it's worth noting that this is a joint UK-US Air Force facility and it appears to have been targeted. We have many joint facilities in Australia with the United States that also could be targeted, and we have had some very troubling recent incidents around base security. It's a reminder of how vulnerable these bases can be and of the need to invest adequately in defending them.
CLARKE: Does it suggest that actions of the US military increases risks to Australian military facilities?
PATERSON: I think the strategic environment is what increases the risk to us. We have rising powers that are investing in military modernisation that we've never seen in peacetime before. We have extremist terrorists going around the world looking for targets, soft and hard, and that just demands investment and taking security seriously.
CLARKE: And your Queensland colleague Henry Pike believes One Nation was trying to convince him to defect at a meeting that took place in January at one of Gina Rinehart's properties. Have you ever been approached about joining One Nation?
PATERSON: No, I haven't been, and I'm not waiting by the phone.
CLARKE: Is there a campaign underway by One Nation to try and pull over sitting MPs?
PATERSON: There may well be, if so, I don't think it'll be very successful. People like Henry are loyal Liberals, and we're here to fight for our country. We know we've got work to do to demonstrate that we've got the plan and the team to change Australia in the way that Australians are very clearly signalling that they want. But I think we're the only ones that can solve this problem.
CLARKE: Senator Paterson, thanks very much for speaking to me this morning.
PATERSON: Thanks for having me.
ENDS