00:00:00
Speaker 1: And we have resident senior fellow and head of AI and security at the Australian Policy Institute, David Rowe, on the line. David, good morning. Thank you for your time.
00:00:09
Speaker 2: G'day, Matthew.
00:00:10
Speaker 1: This is really disturbing. So as I understand it from the reports this morning, and now we know there's been four separate government websites that have been probed by AI. This is just the, what's the word, the machinery, the AI deciding to do this on its own. Is that correct?
00:00:27
Speaker 2: Partly, yes. So it was sent out to gather health data or health funding data, presumably from around the world. What it took upon itself was when it hit against a digital barrier, a kind of firewall between the public and the private sector, the privately held data, it decided that the best way forward was just to go through the firewall, to hack through it effectively. So I think the critical phrase that's been used this morning is Albanese saying that it didn't take no for an answer. That's really important because that is exactly how we're building data. new powerful artificial intelligence models, they are trained to pursue objectives and to pursue them hard and not to give up at the first sign of failure. So that's what's actually happened here. It's a kind of a feature of the new systems, not a bug. So I think the frightening thing, it's less about what was actually compromised here, although obviously that's worrying, but it's more about the trend because these things are getting more powerful, they're getting more independent and autonomous and able to act on their own. And this is going to keep happening and it's going to get more serious without better safeguards.
00:01:39
Speaker 1: Yeah, okay. So what do we need to do? I mean, not give it hacking powers for a start, presumably.
00:01:46
Speaker 2: Well, unfortunately, the hacking powers are the flip side of the coin from being very good at writing code. So it's– unfortunately, what makes them useful in many ways, which is being able to write code for good reasons, also makes them really, really good at hacking. So it's that sort of light and dark side of the same skill. Look, there are– I mean, look, it's– It's really hard to build increasingly powerful models without introducing risks, especially if you give them like sort of higher level goals. In this case, scale the world for health funding data and presumably pull together a whole bunch of information and turn it into a kind of analysis or something like that. That's a pretty high level goal. It's much higher than, say, go to one website, pull some data off and crunch the data for me. The more latitude you give it, the greater the risk that it's going to go about it in some way that's dangerous or damaging.
00:02:46
Speaker 1: You said at the start it was sent out to gather information. Who sent it out? Who programmed it to look for the information? Because surely that's the source we go to.
00:02:56
Speaker 2: Well, it's the open AI researchers in this case. So they were carrying out, it appears, I mean, we're sort of going off the statements that have been released, but it appears that they were testing an internal model. So it's not one that has been released. It's not on your chat GPT at the moment, but it's an internal model. They were testing how good it would be at doing something like, you know, data-driven research on, you know, in this case, health systems. And so they, you know, they, they, put it out on the public internet, although, like I said, it's not released for ordinary punters like you and me to use, but they were testing it to see how good it was. Basically, they were running it through its straps. And so, yes, ultimately, they are the ones who are accountable here.
00:03:38
Speaker 1: All right. So we clearly need in Australia an artificial intelligence capability that works in the national interest to have one of our own. I don't know if that'll stop. It probably won't. Will it stop other AI from overseas trying to get in?
00:03:54
Speaker 2: It could. And look, you can definitely use artificial intelligence for cyber defense as well as cyber attack. And in fact, you can use it for a whole bunch of national security purposes. And we definitely should have that. At the moment, we've still got reasonable access to the best American technology. The government or big companies, they can use American AI services to defend themselves or to do all sorts of things. The big question for Australia is, in another year to five years, are we still going to get access to the best American technology? Because it's getting so good... that it's becoming harder and harder for even a close friend like the United States to share this around with people with even its allies. So there is a big question about whether Australia needs to start developing its own capability. Unfortunately, it's a really hard thing to do. There is literally trillions of dollars in investment going into these companies, and so they are extremely well-resourced and well-funded and good at what they do. It's very hard for a country like Australia to catch up, but we do need to look at how we make sure that we continue to get access, whether it's homegrown or imported, to the best technology.
00:05:06
Speaker 1: Absolutely. David, look, thank you for your time this morning. Resident Senior Fellow, Head of AI and Security at the Australian Policy Institute. Thank you, David. The Deputy Prime Minister, now Acting Prime Minister Richard Miles, has started talking about this at his media conference. We'll have a quick listen.
00:05:23
Speaker 3: I want to say and acknowledge that OpenAI have been working with us very cooperatively here. They have clearly notified us of this and engagement with them has been critical to understanding what has occurred. We are grateful for that. That said, what has occurred here is obviously completely unacceptable and that was made clear in the plainest terms by the Prime Minister in speaking with the CEO of OpenAI over the course of last night. All of this I think we can take as a warning in relation to the development of AI itself. It is so important that this technology, which represents such an enormous opportunity for humanity, is nevertheless being developed in a way where the guardrails, the safeguards, are well ahead of the capability itself. And that's why the Prime Minister, along with other world leaders in New York over the course of this week, has put out a call saying, to ensure that AI is developed in a safe way with guardrails in place. That is very clearly the position of the Australian government as articulated by the Prime Minister. For what it's worth, it's also the position of the CEOs of these tech companies. Over the course of last night, Dario Amodi, Sam Altman and others appeared at the UN to express exactly this same sentiment. We saw Dario Amodi, the CEO of Anthropix, essay in the last couple of weeks going exactly to this point. AI is a transformational technology. It offers so much better benefit for humanity, but it comes with danger. And it is really important that the safeguards and the guardrails are well ahead of of the capability itself, and that is very much the position of the Australian government and the way in which we will be engaging with this technology and seeking to regulate and manage it in an Australian context.
00:07:25
Speaker 4: Thanks, Richard. So I'm here as Minister for Government Services, which the Agency Services Australia sits under. So I thought it might be useful if I can take you through some of the Services Australia information I can give you right now.
00:07:39
Speaker 1: This is Katie Gallagher.
00:07:40
Speaker 4: As up to date as it can be. So on the 10th of September, Services Australia was notified by OpenAI that an AI agent had accessed infrastructure behind the public-facing Medicare statistics reporting service portal, which hosts that publicly available aggregate Medicare and pharmaceutical benefit scheme statistical data that the DPM just mentioned. The agent was undertaking a task by OpenAI to conduct internet-based research into public medicine spending as part of internal capability evaluation. So on the website that it accessed, the Medicare Statistics Reporting Service portal is a standalone website. It's public-facing and it is not in any way related to Medicare in terms of claims, benefits, payments, processing individual.
00:08:36
Speaker 2: Information.
00:08:37
Speaker 4: So it's a completely different system and the two shouldn't be conflated.
00:08:42
Speaker 3: It's a public website.
00:08:43
Speaker 4: It's most often used by researchers and academics who get that aggregated data about benefit statistics, prescribing statistics to use in their own research.
00:08:54
Speaker 1: All right, Katie Gallagher there saying essentially... As she and the government have said, private information wasn't breached, at least as far as we know so far. But the information, and as David Rowe said too, they were searching up public health information. But where could it go next? That's the scary bit. More on this to come, and you'll hear updates on this story throughout the day here on 5AA News.
Speaker 1: And we have resident senior fellow and head of AI and security at the Australian Policy Institute, David Rowe, on the line. David, good morning. Thank you for your time.
00:00:09
Speaker 2: G'day, Matthew.
00:00:10
Speaker 1: This is really disturbing. So as I understand it from the reports this morning, and now we know there's been four separate government websites that have been probed by AI. This is just the, what's the word, the machinery, the AI deciding to do this on its own. Is that correct?
00:00:27
Speaker 2: Partly, yes. So it was sent out to gather health data or health funding data, presumably from around the world. What it took upon itself was when it hit against a digital barrier, a kind of firewall between the public and the private sector, the privately held data, it decided that the best way forward was just to go through the firewall, to hack through it effectively. So I think the critical phrase that's been used this morning is Albanese saying that it didn't take no for an answer. That's really important because that is exactly how we're building data. new powerful artificial intelligence models, they are trained to pursue objectives and to pursue them hard and not to give up at the first sign of failure. So that's what's actually happened here. It's a kind of a feature of the new systems, not a bug. So I think the frightening thing, it's less about what was actually compromised here, although obviously that's worrying, but it's more about the trend because these things are getting more powerful, they're getting more independent and autonomous and able to act on their own. And this is going to keep happening and it's going to get more serious without better safeguards.
00:01:39
Speaker 1: Yeah, okay. So what do we need to do? I mean, not give it hacking powers for a start, presumably.
00:01:46
Speaker 2: Well, unfortunately, the hacking powers are the flip side of the coin from being very good at writing code. So it's– unfortunately, what makes them useful in many ways, which is being able to write code for good reasons, also makes them really, really good at hacking. So it's that sort of light and dark side of the same skill. Look, there are– I mean, look, it's– It's really hard to build increasingly powerful models without introducing risks, especially if you give them like sort of higher level goals. In this case, scale the world for health funding data and presumably pull together a whole bunch of information and turn it into a kind of analysis or something like that. That's a pretty high level goal. It's much higher than, say, go to one website, pull some data off and crunch the data for me. The more latitude you give it, the greater the risk that it's going to go about it in some way that's dangerous or damaging.
00:02:46
Speaker 1: You said at the start it was sent out to gather information. Who sent it out? Who programmed it to look for the information? Because surely that's the source we go to.
00:02:56
Speaker 2: Well, it's the open AI researchers in this case. So they were carrying out, it appears, I mean, we're sort of going off the statements that have been released, but it appears that they were testing an internal model. So it's not one that has been released. It's not on your chat GPT at the moment, but it's an internal model. They were testing how good it would be at doing something like, you know, data-driven research on, you know, in this case, health systems. And so they, you know, they, they, put it out on the public internet, although, like I said, it's not released for ordinary punters like you and me to use, but they were testing it to see how good it was. Basically, they were running it through its straps. And so, yes, ultimately, they are the ones who are accountable here.
00:03:38
Speaker 1: All right. So we clearly need in Australia an artificial intelligence capability that works in the national interest to have one of our own. I don't know if that'll stop. It probably won't. Will it stop other AI from overseas trying to get in?
00:03:54
Speaker 2: It could. And look, you can definitely use artificial intelligence for cyber defense as well as cyber attack. And in fact, you can use it for a whole bunch of national security purposes. And we definitely should have that. At the moment, we've still got reasonable access to the best American technology. The government or big companies, they can use American AI services to defend themselves or to do all sorts of things. The big question for Australia is, in another year to five years, are we still going to get access to the best American technology? Because it's getting so good... that it's becoming harder and harder for even a close friend like the United States to share this around with people with even its allies. So there is a big question about whether Australia needs to start developing its own capability. Unfortunately, it's a really hard thing to do. There is literally trillions of dollars in investment going into these companies, and so they are extremely well-resourced and well-funded and good at what they do. It's very hard for a country like Australia to catch up, but we do need to look at how we make sure that we continue to get access, whether it's homegrown or imported, to the best technology.
00:05:06
Speaker 1: Absolutely. David, look, thank you for your time this morning. Resident Senior Fellow, Head of AI and Security at the Australian Policy Institute. Thank you, David. The Deputy Prime Minister, now Acting Prime Minister Richard Miles, has started talking about this at his media conference. We'll have a quick listen.
00:05:23
Speaker 3: I want to say and acknowledge that OpenAI have been working with us very cooperatively here. They have clearly notified us of this and engagement with them has been critical to understanding what has occurred. We are grateful for that. That said, what has occurred here is obviously completely unacceptable and that was made clear in the plainest terms by the Prime Minister in speaking with the CEO of OpenAI over the course of last night. All of this I think we can take as a warning in relation to the development of AI itself. It is so important that this technology, which represents such an enormous opportunity for humanity, is nevertheless being developed in a way where the guardrails, the safeguards, are well ahead of the capability itself. And that's why the Prime Minister, along with other world leaders in New York over the course of this week, has put out a call saying, to ensure that AI is developed in a safe way with guardrails in place. That is very clearly the position of the Australian government as articulated by the Prime Minister. For what it's worth, it's also the position of the CEOs of these tech companies. Over the course of last night, Dario Amodi, Sam Altman and others appeared at the UN to express exactly this same sentiment. We saw Dario Amodi, the CEO of Anthropix, essay in the last couple of weeks going exactly to this point. AI is a transformational technology. It offers so much better benefit for humanity, but it comes with danger. And it is really important that the safeguards and the guardrails are well ahead of of the capability itself, and that is very much the position of the Australian government and the way in which we will be engaging with this technology and seeking to regulate and manage it in an Australian context.
00:07:25
Speaker 4: Thanks, Richard. So I'm here as Minister for Government Services, which the Agency Services Australia sits under. So I thought it might be useful if I can take you through some of the Services Australia information I can give you right now.
00:07:39
Speaker 1: This is Katie Gallagher.
00:07:40
Speaker 4: As up to date as it can be. So on the 10th of September, Services Australia was notified by OpenAI that an AI agent had accessed infrastructure behind the public-facing Medicare statistics reporting service portal, which hosts that publicly available aggregate Medicare and pharmaceutical benefit scheme statistical data that the DPM just mentioned. The agent was undertaking a task by OpenAI to conduct internet-based research into public medicine spending as part of internal capability evaluation. So on the website that it accessed, the Medicare Statistics Reporting Service portal is a standalone website. It's public-facing and it is not in any way related to Medicare in terms of claims, benefits, payments, processing individual.
00:08:36
Speaker 2: Information.
00:08:37
Speaker 4: So it's a completely different system and the two shouldn't be conflated.
00:08:42
Speaker 3: It's a public website.
00:08:43
Speaker 4: It's most often used by researchers and academics who get that aggregated data about benefit statistics, prescribing statistics to use in their own research.
00:08:54
Speaker 1: All right, Katie Gallagher there saying essentially... As she and the government have said, private information wasn't breached, at least as far as we know so far. But the information, and as David Rowe said too, they were searching up public health information. But where could it go next? That's the scary bit. More on this to come, and you'll hear updates on this story throughout the day here on 5AA News.
