CamFeed

[0.24s] Speaker 2: A major tech company claims that one of their powerful AI models has broken out of its restraints and hacked into another company’s computers. [9.04s] Speaker 2: OpenAI says it was testing one of its products when it noticed that the AI model had forced its way onto the open web as it tried to carry out a challenge it had been given. [19.68s] Speaker 2: The company says this autonomous capability is an example of the safety concerns that the AI industry has to prepare for. [27.28s] Speaker 2: Cam Wilson is the ABC’s national AI technology reporter and is with me in the studio. [31.76s] Speaker 2: Cam, it's an incredible story. [33.28s] Speaker 2: Can you talk us through exactly what happened? [35.44s] Cam Wilson: Yeah, so last week we had an announcement from an AI company saying that someone had hacked into our computers, our servers, and we’d noticed this, and we believe it was actually an autonomous AI agent, which is a way of saying it wasn't a person who was doing this, it was an AI thinking by itself. [52.24s] Cam Wilson: People saw that and was like, oh that’s curious, but didn’t know much about it. [56.4s] CamWilson: Today, OpenAI, which is the maker of ChatGPT, the AI company that many people know about, disclosed that it was responsible for that, and specifically that its AI models had hacked into this other company's servers, not being asked to do that. [74.72s] Cam Wilson: It had decided to do it on its own, and as a result, and it got out of these kind of restraints that have been given, showing why they have been talking a lot about how carefully we need to take this safety issue. [86.48s] Speaker 2: So is this an example or a signal that AI is just becoming too clever for us? [91.76s] Speaker 2: And did it know that what it was doing and hacking into another company was wrong? [95.6s] Cam Wilson: Yeah, so the details of what exactly happened that OpenAI shared was it had an internal testing environment. [102.72s] Cam Wilson: So it was looking at how a current model that’s out that anyone can use, but also an unreleased model that they say is even more powerful, would react given certain tasks, a kind of benchmarks. [114.56s] Cam Wilson: One of these benchmarks is saying, Hey, can you do this challenge? [117.36s] Cam Wilson: Can you try and do something? [118.8s] Cam Wilson: And this environment that it was in, so this kind of sandbox was restricted from the open internet because they’re saying we're trying to do this responsibly, we're putting it through its motions. [129.0s] Cam Wilson: But it had a very narrow path out of it, which was saying, hey, if you want to use these tools which are on the internet, you can use this very, very narrow path. [137.0s] Cam Wilson: It’s kind of like saying you can access like one internet website. [139.96s] Cam Wilson: What the AI decided to do is, given a task, it decided instead of trying to go through that task and um uh complete it as they thought it might, it found a way to kind of force its way through the small opening, work through other computers on the open AI's network, get onto the open internet, and to go to another company’s servers, which it believed had the answer to the challenge that it was trying to complete. [167.64s] Cam Wilson: So it was saying, I want to do this challenge, you told me to do it, I'm gonna do it, but the way that it went about it, which involves some pretty sophisticated like cyber intrusions, things that they did not expect, and in fact thought that they had actually stopped being possible, it did all that in the task in the kind of attempt to complete this task. [186.92s] Cam Wilson: So, how serious was the breach, uh Cam and could Hugging Face use as the company that was hacked be affected by it? [193.64s] Cam Wilson: Yeah, so Huging Face says it wasn’t a serious breach, and it said that you know it was just some of its testing environments. [198.36s] Cam Wilson: So essentially saying, you know, these are not the kind of you know really secure stuff that we have, and and it wasn't probably the lockdown of some of its other stuff. [205.48s] Cam Wilson: But you know, both companies are taking it exceedingly seriously because one of the things that they talk a lot about in the AI industry is not just like what happens if someone nefarious says, I’ve got access to these things and I want you to do this thing and are intentional about it, but what they call alignment, which is this idea of how does AI act, all the little decisions it makes when trying to do something that you might not necessarily say to do. [230.56s] Cam Wilson: For example, you know, do this challenge, you might not say do this step, this step, this step. [235.44s] CamWilson: And in doing so, like they they try to say, how can we train these AIs so that they act ethically and among like, you know, within our expectations, based on uh uh like before we even know what they're doing. [248.48s] Cam Wilson: And this is an example of like seemingly them doing something incredibly no potentially damaging that has clearly sent uh like shock waves through the industry. [257.92s] Cam Wilson: Thanks for explaining it all, Cam. [259.68s] Cam Wilson: Thank you.

The Prime Minister is set to make a significant speech on artificial intelligence this week. His second on AI in just over a fortnight is encouraging a lot of speculation about whether the government will provide more details about its expectations of data centres and AI infrastructure developers. The ABC's national AI reporter is Cam Wilson. Hi, Cam. Is it looking like the Prime Minister is about to announce a shift in the government's AI policy? So we're hearing that there will be a big announcement made later this week. There was a noticeable shift in how the Prime Minister was talking about AI and the fact that he was talking about it altogether a few weeks ago when he gave a speech, which was talking to an economic forum, and as part of that, he was talking about many of the other governments' other initiatives, such as their push for renewable energy, and as part of that, he threw AI into the mix. It wasn't a big part of the speech, but it was very early on and kind of prominently put. And he was speaking about how in Australia, you know, we have to deal with these factors that are changing all around the world. Technology is changing, the world is changing, and we have to respond to it and make the best of it. We can't just let the world, you know, change us. We've got to respond and you know do what we can. So, as part of that, people have kind of believe, based on everything else that is happening, that they are looking to announce some decisions on some of the policy issues that will affect whether AI companies are looking to make major investments in Australia. Okay, so there has been a bit of a pushback overseas, and now we're seeing it in Australia in terms of the rise of data centres, the way that they're using scarce resources, power and water, in addition to some of the localities in which they're based. They're popping up in some suburban areas. Is this speech about also assuring or making the case to Australians about why we need to have AI based in Australia? It looks like he's going to be trying to make that case. It's made a really big effort to broadcast the fact that it is hard on big tech, lots of cracking down on big tech. And this, the data center interest that has kind of come out of nowhere, is another area in which they are seeing that there is a lot of public interest. When I talk to politicians, people who are in the industry and policy, they are all saying that the issue of data centers has really become this enormous focus for public attention in a way that they are completely surprised. In fact, when I spoke to people from the data center industry in particular, they were saying two years ago no one cared if we put up a data center, and now we are in the middle of this like political firestorm. So the government no doubt sees this as happening already in Australia. It sees America, where to be fair, it is a bit of a different issue. The data centers in the US are much bigger, they have a different infrastructure setup. But of course, you know, the basic dynamics are the same, that these are these, you know, new buildings that take up resources that compete in the local community to, you know, power these systems that some people were very fond of. Some people have big reservations about. They are seeing this, understanding that this is an issue that they need to address. And I think they're seeing it as an opportunity to establish their kind of bona fides on this topic, to say, yes, data centers can come to Australia, yes, you can build them here, but in return, we have these demands, and some of the demands that they've kind of foreshadowed are things about you know economic fairness, opportunities, you know, making sure that they are continuing to be based on renewables and helping, not hurting that transition. A lot of things that the government has already been interested on that it now sees as a data center as an opportunity to then mark their stamp. And so is the government also seeking here to talk directly to big tech and the investment sector, given the huge capital that is required for the AI sector to advance in Australia? And what is the message if that is indeed the case? Yeah, so I mean, we are seeing that many of these hyperscalers, which is the big tech companies that are looking to build data centers around the world, not so much just the ones that are used to run these services, that's kind of called inference. So, you know, if you use an AI chatbot, it's kind of like those are the data centers that are kind of smaller, typically need to be placed in metro areas so they're close to the people who are using them. Then you've got these bigger ones that are used for the training. These are you know very, very much like power consuming, resource-intensive, much bigger in scale that many of these companies are looking to place outside of the US for a variety of reasons. Australia has emerged as one of the places that some of these companies are looking to potentially build these. It's a massive investment, but they also see Australia as an attractive place to do it for a variety of reasons. We're politically stable, we've got, you know, a um we've got good education here, we've got you know good infrastructure, renewable energy. For all these reasons, they are thinking about doing it here, but one of the big issues, a major hurdle, is this issue of copyright, this idea that training AI in Australia may not be legal under copyright. They're kind of looking for an answer on that. And I think what we're seeing from the government is maybe we'll see something soon on copyright, but they're also being very clear that before we even have that conversation, we want to set down the rules for Thank you even coming here and thinking about building these data centers. you. Ham Wilson will know more later this week. Thank you.

Speaker 2: Now, South Australian police have also confirmed today that a report suggesting there was a failed call to 000 due to the Telstra outage that ended in a death are incorrect. Speaker 2: On Wednesday night, Liberal Senator Karen Little posted on social media claiming her office had received a report of a death following an apparent failure to connect to Triple Zero. Speaker 2: She stated an elderly South Australian person had died. Speaker 2: But today, South Australian Police Commissioner Grant Stevens says a thorough investigation has been carried out, and it found that calls made to Triple Zero using a Telstra mobile phone for this particular family did connect, and an ambulance was requested. Speaker 1: This is a tragedy where a family has lost a loved one. Speaker 1: And in the ordinary course of events, they would not have had such significant involvement with the South Australia police, and they would not, this particular person's passing would not have been the subject of such significant scrutiny. Speaker 2: You’ve today reported that Telstra was repeatedly warned by government agencies about this sort of outage occurring. Speaker 2: What exactly were they told? Speaker 1: Yeah, that’s right. Speaker 1: So Telstra, in the years and months leading up to this week's outage, was told that the system that suffered the fault this week was vulnerable to a kind of error like this that could cause some massive issues for the telco. Speaker 1: The system at hand, the one that we’re talking about that’s been identified by Telstra, is its timing system and its dependency on GPS satellites. Speaker 1: Most people think of GPS, they think of location, but another really important function that they serve is also telling the time very accurately to computers all around the world that are used in a lot of infrastructure, including telecommunications. Speaker 1: So, for example, telecommunication networks use this very specific timing to be able to coordinate when one cell tower talks and the other one listens, and also as part of their security, because when a cell tower receives a signal and that it believes it is from another cell tower, the way that it figures out if that's a legitimately from a cell tower, not someone else trying to interfere, is it looks at a security certificate, which is essentially a way of knowing that the signal comes from who it says it is. Speaker 1: That depends on also the time as well. Speaker 2: So when you had the error that happened this week, which reportedly convinced Telstra systems that it wasn’t 2026 but actually 2006, that caused this error, it messed up with these settings that caused the problems cascading across the network. Speaker 2: Now, um, this is something that had been brought up by a government agency, the federal government's critical infrastructure security agency and experts directly to Telstra this year. Speaker 2: And Vicky Brady was asked about uh this reporting today. Speaker 2: Well, I think we need to get underneath exactly what was part of that conversation as I understand it right now. Speaker 2: It was a broad range of things um spoken about, and as I Speaker 2: mentioned, these timing synchronization systems are well known to be critical to a mobile network so now Speaker 2: Cam you mention Australia's government run cyber and infrastructure security centre repeatedly raised timekeeping as a major concern for critical infrastructure providers so aside from raising the issue can the government actually compel or force telcos to act on such a concern Speaker 2: agency that put out those awards is not one that has the powers to do anything but Telstra as a company and a telecommunications provider that has obligations under legislation and rules for critical infrastructure in Australia could be asked to do certain kinds of things. Speaker 2: There was a new rule brought in last year that requires these networks to prepare for all kind of potential hazards to the network presumably having some kind of backup for the the timekeeping system that was affected by this error or just generally having a kind of ability to test and have a resilience when there are issues to its timing system altogether the act also gives the minister in the the the federal government's critical infrastructure security agency and experts directly to Telstra this year. Speaker 2: Now, um, this is something that had been brought up by a government agency, the federal government’s critical infrastructure security center repeatedly raised timekeeping as a major concern for critical infrastructure providers so aside from raising the issue can the government actually compel or force telcos to act on such a concern Speaker 1: agency that put out those awards is not one that has the powers to do anything but Telstra as a company and a telecommunications provider that has obligations under legislation and rules for critical infrastructure in Australia could be asked to do certain kinds of things. Speaker 1: There was a new rule brought in last year that requires these networks to prepare for all kind of potential hazards to the network presumably having some kind of backup for the the timekeeping system that was affected by this error or just generally having a kind of ability to test and have a resilience when there are issues to its timing system altogether the act also gives the minister in the the federal government's critical infrastructure security agency and experts directly to Telstra this year. Speaker 2: Cam Wilson thank you for your time. Speaker 1: Thank you.

Well, more now on the alleged warnings that Telstra is supposed to have been given about its system. ABC National AI Technology reporter Cam Wilson has been following that story and is with me in the studio. Cam, you're also at that media conference with Telstra this morning. What did we learn about this supposed error? So we got a bit more information about the error, but what we heard mostly about today was their response to the error. Telstra's CEO and CFO were both out there talking about how they had heard about it very soon after it happened and how their teams had kicked into response very quickly. But despite that, they said this error continued to cascade across the network, ultimately bringing it down. So, you know, this kind of error is a error that they acknowledged was a known risk, or in fact, well known within the industry, is what I think that they said. But we still did not have much information about how it got to this point. And you reported this morning, Cam that Telstra had been warned in the months, even years leading up to the outage about the error. What warnings had Telstra received before this happened? Yeah, so Telstra acknowledged today that they had received these warnings from government agencies and academics. There is a federal government agency that deals with the security of critical infrastructure, which includes things like telecommunications companies. And they had repeatedly publicly warned that there is a potential issue with the timing system as a vulnerability. Now, timing is something that probably, you know, most people haven't thought much about when we think about the satellites that do timing, probably think about them more in relation to the fact that they also do location. These are GPS satellites that tell us where we are, they help us navigate, but they also serve this important function of telling computers very precisely what time it is. That is important for all kinds of infrastructure, including telcos, which need that information so that their cell towers can talk to each other. You know, they need to know like when one cell tower is talking, the other one's listening, and also as part of their security. When a cell tower receives a signal, it has what's known as a security certificate. This is supposed to be the thing that vouches for the fact that what's incoming is from who it says it's from. Those have expiry dates, and unfortunately, we're told this issue affected their conception of what time it was, which meant that you know these security certificates then were no longer working, which caused the issue. So Telstra said we're aware of this issue. Um government agencies had said, you know, you really need to prepare for this, and even academics have been increasingly sounding the alarm about timing systems as a potential point of failure across telcos and other infrastructure. One who spoke to me said that they had personally gone to Telstra this year saying that they believed it was an issue, but obviously, you know, this has still ended up happening. Let's hear from CEO Becky Brady now about the error. This issue of time synchronization in our network, that's when the first issue occurred. That does take time to propagate across a very large network. Our teams, our technical teams, were working on it almost immediately. So they acknowledged this error that they had known about things like this, but it was also very interesting what they were evasive on during the press conference. Um, they wouldn't say specifically that they had a backup system for the timing system. They just said, you know, generically that they had backups and resilience in their system. So it'll be interesting to see if they do give us some more details on that. And the other thing is they wouldn't actually say that in their legally required risk management plan, had they actually identified this type of error as something that could affect them and prepared strategies to be able to respond to that. That's something they have to do. The government actually has that information. I expect that we'll hear more about this because, of course, this critical infrastructure is something that the government has been putting a lot of pressure on. It's put um it's passed new rules recently saying you need to prepare for these kinds of errors and risks and vulnerabilities. This vulnerability has happened. They didn't seem that prepared. So I guess we'll see how what they had said and whether they had lived up to that or they failed to do the things that they promised to do to address this.

[0.08s] Speaker 2: Meanwhile, Telstra is being questioned as to why it wasn’t prepared for the issue blamed for the network outage after the ABC revealed the federal government's critical infrastructure agency had repeatedly warned telco providers of the potential vulnerability. [16.08s] Speaker 2: A national AI reporter Cameron Wilson broke this story, and he joins me now with more. [21.52s] Speaker 2: Hi there, Ken. [22.32s] Speaker 2: So just take us through what we know about this error. [26.08s] Cam Wilson: So we still don’t know everything, but Telstra has been giving us more information in the days since the outage earlier this week. [32.96s] CamWilson: They've pointed to a software error that happened as a result of an update and have linked it to their GPS system that is used as part of the timekeeping across the network. [49.52s] Cam Wilson: But another important but less known function of GPS satellites is also telling very accurately computer systems what time it is. [57.6s] CamWilson: And timing is a very important kind of system that’s needed across a whole array of very important critical infrastructure, including uh telecommunication networks. [69.12s] Cam Wilson: The role that it fills is that things from as simple as cell towers need to talk to each other, they need to know when to talk, when to listen, and also to matters of security and authorization. [81.36s] CamWilson: So, for example, if I'm a cell tower and I get a message from someone but I don’t know who it is, I’m told it’s another cell tower. [88.24s] Cam Wilson: Um, one of the ways I do that is by looking at their security certificate. [92.16s] CamWilson: And those security certificates rely on uh the time they often have a time and date expiration. [99.36s] CamWilson: Uh so what we're told happened and what has also been reported is that all of a sudden Telstra’s GPS timekeeping system believed that it was not 2026 but 2006, and that created errors that cascaded across the network to cause the outage, but of course, then cause these issues across the country. [118.16s] Speaker 2: And so, what were the warnings that Telstra received? [121.88s] CamWilson: So timing is one of those things that has been coming, has been raised repeatedly by agencies who are looking at Australia’s critical infrastructure and also by experts who've been studying it. [134.76s] Cam Wilson: The Australia’s Critical Infrastructure Security Agency had repeatedly warned critical infrastructure providers, especially telecommunication providers publicly over the last few years that timing was something that they needed to make sure that they were not vulnerable to. [149.96s] CamWilson: They identified it as a potential area that needed to have resilience. [154.76s] Cam Wilson: And so broadly they said, you know, they were looking in particular at these satellite based systems, but also, you know, there are other options for timing. [162.92s] CamWilson: You can have backup systems like atomic clocks. [165.56s] Cam Wilson: Telstra did say that it has a backup system, but that was not able to prevent the error. [170.44s] CamWilson: Um the question of the timing system and preparing for an outright failure is something that can be tested for. [187.96s] Cam Wilson: Of course, it is a very important part of these systems. [199.24s] Cam Wilson: The experts I spoke to spoke about having a variety of ways to prepare for errors like this, whether it is having the backups, doing more testing of what happens uh when something goes wrong to your major provider if it is the satellite provider, and also just generally running drills and even potentially in Australia, having something like the UK just launched, a national timing centre that would not only provide a source of timing for Australian companies and infrastructure, but would also kind of be responsible for keeping in touch and coordinating across sectors in you know preparing for these kinds of failures. [214.92s] CamWilson: The federal government late last year did create a new legal obligation for telecommunication providers to have a risk management plan for all kinds of errors, and one of the specific errors that wasn’t included in this was a failure to these satellite systems. [219.24s] Cam Wilson: So we know that they were tasked with doing it. [229.76s] CamWilson: The Department of Home Affairs confirmed to the ABC that they had received that plan from Telstra. [239.24s] CamWilson: We don’t know what's in it. [241.92s] Cam Wilson: Obviously, you don’t want to just go handing out a document with all of your potential errors. [247.88s] Cam Wilson: That seems like something that's probably not good to have out there under the public. [248.96s] CamWilson: Um so I guess that will be one of the areas of inquiry going forward. [250.12s] Cam Wilson: How did they prepare for this if they identified it as a risk? [253.76s] Speaker 2: Okay, National AI and Technology Reporter CameronWilson there with those details. [254.84s] Speaker 2: Thank [255.00s] Cam Wilson: you. Thank you.

The ABC has told staff it’s embracing artificial intelligence through the organization. It includes rolling out tools to staff, hiring for AI-specific roles, and running a pilot program of using AI to help turn regional radio bulletins into online stories. The broadcaster says this technology can help journalists free up time for reporting. Experts and journalists warn misuse could damage trust in one of Australia’s most trusted news organizations. Cam Wilson is the ABC's national AI technology reporter and joins us here in the studio. The ABC emailed staff late last month, saying they’ve got a range of initiatives that we’re launching with the overarching idea that generative AI was created with ChatGPT at the end of 2022. It’s happening, and it’s a technology that we should try to embrace to get the benefits while mitigating the risks. The ABC is cautiously welcomed by unions and experts who said you know, if you do it as you say you're going to do it responsibly, then you could see this being something that we would like to see. Essentially, there’s a balancing act. The ABC initially stated its plans for rollout was cautious, but the AI principles put out in 2024 said that any use of AI needs to be disclosed. The ABC remains a little behind other public broadcasters around the world, and it's not practical to say every time AI use is used. It’s in all kinds of things now. AI touches everything now, so there are many places where AI uses. There are guidelines saying that if you're using this tool, you think it will affect how this content looks and feels to audiences, you need to tell them that. The ABC has also made a new guideline stating that the ABC needs to disclose any use of AI in the production of content. It’s not transparent because the initial principles put out in 2024 said that any use of AI needs to be disclosed. The ABC is glad I'm not the person who has to come up with these principles themselves.

Ian Gillespie reporting there from Brisbane. Dozens of government bodies missed deadlines for reporting how they use artificial intelligence and nominate a single person to be in charge of its use. It was the first test of the government's plan to make agencies responsible for policing their own AI use. And while the departments have since corrected their mistakes, the lapses have prompted questions about Australia's broader approach to regulating AI. Is national AI technology reporter Cam Wilson. Last year Australia was at a crossroads. We were deciding which way we're going to go with AI regulation. One way was the EU style, where they have a central AI Act with rules in it that cover the technology's use across the whole country. The other pathway that we ended up going down was instead of doing that, we decided to give the responsibility of regulating AI use within various sectors to the government agencies that were responsible for regulating that sector overall. So instead of these one rules, you know, from the central organization, it was given to everyone to say you're the experts in your field, figure out how to use AI. Now that's obviously a different approach, and you know they bring that expertise, but there has been questions about well, you know, how are they going to understand this new technology? And so it was worth looking at how in the past they have regulated their own use of AI themselves to see just how good they are at that task. And so what did you or the ABC find about how agencies have handled this? Yeah, so around the same time, uh, the uh government agency that's responsible for administering technology across the government, which is the digital tech uh tech no digital transformation agency, um, they create a policy saying government agencies need to build up trust in themselves and how they're using this new technology. So they tasked all the agencies with disclosing how they use AI and also picking someone who's literally responsible for it, making sure that someone actually answers for it at the end of the day. And what we found was that uh dozens of these agencies out of about 90 odd uh had actually failed to do one or the other um by the first deadline that they've kind of set up in it. So this first task of saying, you know, not even like regulating outside of themselves, but literally how are you using it, they didn't meet it. Now that's not to say that they aren't using it in a safe way, but they weren't doing a great job at kind of meeting these basic expectations of explaining that, and so that makes it very hard from the outside to know how they're using it and to build that trust in their use of it. Are there possible concerning implications from this? Yeah, I mean, I think the thing that you would just like assume is that you know, if you'll say we're we've got to go one way rather than the other, and you look at how they've kind of done that similar task in the past, um, it wasn't necessarily instilling a whole lot of confidence. That being said, they have you know um met those expectations, all of them now, so they have you know remedied that, and there are new expectations, broaden expectations to give us more detail about how they are using it within the agencies. So we'll see how they go on that. Some of those deadlines are still to come, but obviously, this is an area that's so important. It's this uh you know type of technology that is very hyped, Australians don't generally have a lot of trust in it, and it's changing so quickly, and it's very technical. So it's something that people are paying a lot of attention to. So I guess we'll see how they go on the next set of deadlines. And you'll be on it. I will be. Okay, we'll leave there, Ken Wilson. Thank you.

Speaker 2: Last year, Australian government agencies had an early stumble at self-policing their own use of artificial intelligence. They were supposed to disclose how they were using the technology and pick one person to be accountable for its use. But dozens of them failed to do so by their deadline, prompting questions about the government's broader plans to leave AI rulemaking up to each agency. Cam Wilson is the ABC’s national AI technology reporter and is with me in the studio now. Good to see you again, Cam. So, firstly, can you tell us about how the Australian government is approaching making rules around how to use AI? Yeah, sure. So last year the Australian government was at a fork in the road. Initially, it had decided to go with and was proposing to go with an EU style regulation that had a single AI Act and these broad rules in a law that would go across the whole country, everything. But after the 2025 federal election, instead they decided to change that and go with the style that left the regulation for AI up to each of the government agencies, tasking them with looking at their own sector and saying, here's how you should use the technology, including self-poicing They were supposed to put out uh AI transparency statements that said, here’s the technology we’re using, here’s what we’re using it to do, here’s how we’re making sure that it is done in a safe way. Um, and also some of them failed to nominate that one person who is supposed to be the person responsible for using the machine. You know, only necessarily that some of them didn't do these two tasks. That being said, without knowing how they’re using it, disclosing it, without knowing who’s responsible, it does make it very hard for us outside to you know feel comfortable and trust in the use of it within government. And what does this mean for their broader plans then? Well, I mean, when you think about it, you know, this decision to give all the agencies individually the task of dealing with this technology. They all have plenty on their plate at the moment, and AI, this new industry and new technology that is constantly changing, is extremely technical and complex. Um, you know, that one of the challenges was that by giving it to all the agencies individually, it meant that maybe it wouldn't be top of their to-do list. Um I should say that I think all the government agencies are now compliant with those basic rules, and the government has now added new rules, and so we’re actually getting more information and more disclosure. Thank you, Cam.

Ian Verender there. Well, last year Australia changed course on its approach to artificial intelligence. Instead of a EU style with a single AI Act and countrywide rules, the federal government chose to task its agencies with coming up with their own rules, including for how they would use the technology within the public sector. But when it came to their first test of self-policing, a majority of government agencies failed. With more, I'm joined by AI reporter Cam Wilson in Sydney. Hi there. Cam, can you just tell us a little bit about the government's early stumbles? Yeah, so one of the first things that government agencies had to do when they were considering talking about using their AI in government was disclose exactly how they were using it and how they were thinking about using it responsibly. This was a policy that was created by the government agency, the Digital Transformation Agency, which is tasked with making sure the government is using technology well and responsibly. It said you have six months, you all uh all these agencies, 94 of them, you've got to tell us how you're using it, how you're thinking about it. And they also was told you have to actually make one person ultimately responsible for it, an accountability officer. Um that six-month deadline came up, and what they found and what we're reporting is that the majority of agencies uh did not actually meet the deadline for those two pretty basic requirements. Um there's a variety of reasons. Some of them got in late, some of them didn't think that it applied to them. Uh honestly, sometimes there were a kind of utopia style areas where um one of the agencies was reportedly not compliant, but when the digital transformation agency looked into it, they realized that their spam filter had actually caught the notice of the fact that they had done this thing. So actually they had they were compliant, it just didn't make it to the right inbox. So yeah, you can see all the ways in which that they are had an early stumbles when it came to this. So then how are they doing now? They've improved since then. This kind of this deadline was uh early in 2025, and since then, all these agencies are now compliant. You know, they they got the the notice and they've fixed it up and they are now all sharing how they're using um AI and they have that officer. Now, when I spoke to an expert, they said the information in these um transparency notices varied. Some were better, some were worse. And this expert cautioned saying that, well, just because these agencies haven't necessarily done a good job of letting us know how they're using AI, doesn't mean they're not thinking about it responsibly, but of course, you know, letting people know about it, being transparent is important to have that awareness to be able to know and to to you know analyze if they are using it well. And so, you know, the expert that I spoke to said this um this still left a little bit wanting. So then what does this mean for Australia's broader approach when it comes to making rules for AI? Yeah, as you mentioned before, um, there was this fork in the road last year where the Australian government decided instead of having a kind of single AI act and maybe even having potentially an AI regulator and having these country to ride country ride wide rules for AI, they instead tasked all the agencies with policing their own sections of the of the country and and themselves. Um I think that this shows that there are challenges with adding another thing to the plates of all these government agencies. Um, the 94 of them have a lot to do. Uh, it gives them another responsibility, and also something that while they're probably the experts on their sector, aren't necessarily the experts on AI. Now, there is some assistance being given by the government with the AI Safety Institute and other things, but you know, there's a lot of things going on right now. AI is this new uh new kind of boom, so much is changing. Um, giving this task to all the agencies does have its potential um issues, and we saw this in this early test. Thank Cam Wilson, thank you very much. you.