GLEESON, Dr Kath, Deputy Electoral Commissioner, Australian Electoral Commission HOUGHTON, Ms Rachel, Chief Operating Officer, Australian Electoral Commission POPE, Mr Jeff, APM, Australian Electoral Commissioner, Australian Electoral Commission [19:11] CHAIR: Welcome. I understand that information on parliamentary privilege and the protection of witnesses has been provided to you. I invite you to make a short opening statement, at the conclusion of which we'll get underway with committee members' questions. Mr Pope : Thank you, Chair, but I don't actually have an opening statement to make. We've made our submission, and I'm happy to go straight into questions and conversation. CHAIR: Excellent; that sounds great. I'll kick off. The integrity of our electoral system is an area of keen interest not just for parliamentarians but for the public more broadly. We've got one of the best electoral systems in the world, and the potential impact of this technology—the impact it may have, could have or is having—on our electoral integrity is of high importance. Could you explain for the committee, or explore or reflect on, what the AEC is currently doing to ensure public confidence in electoral integrity and trust in our democratic process? How are we looking at that at the moment? Could you expand on that? Mr Pope : Electoral integrity, obviously, is critical to the Australian Electoral Commission but not just to us; it's also critical to the credibility and the trust placed in the electoral result and also, therefore, the trust that's placed in parliament and the authority that parliament carries post the election. It's a critical currency, if you like, underpinned by trust and confidence by voters and political participants, so it's obviously something we take incredibly seriously. I might broadly answer your question two ways, and then we can go into a bit more detail if you like. Perhaps internally, within the AEC, to maintain that trust we are taking a very careful and cautious take-up of artificial intelligence. Obviously, we're compliant with the Electoral Act. The election and associated processes are manual. We think that manual processes provide trust and transparency. We'll continue to deliver elections in accordance with the Electoral Act, of course. Any changes to that are a matter for parliament. We want to have a focus on compliance with the legislation. We want a cautious take-up of artificial intelligence; therefore, it'll be a slow and steady approach internally for us. We're putting the foundations in place before we consider the extent to which we might be able to leverage some artificial intelligence, which I suspect will mainly be back-of-house efficiency and administrative support. We have used artificial intelligence a little bit in our IT area to assist our developers and coders, and we've found that quite successful. I'd just like to emphasise it's a careful and cautious take-up. In terms of what we're doing externally to try and manage electoral integrity, our options are somewhat limited—limited by the electoral act, of course. We regulate authorisations of political communications, so, where a communication requires an authorisation, whether or not it's generated by AI, we will regulate that. So, if there is AI generated political communication, that's something we'll absolutely look at. We have section 329 of the electoral act—if that AI communication is a misleading and deceptive publication, that could be a criminal offence, but it is interpreted very narrowly. It only applies to the communication around the casting of a vote and it only applies at election time. So it has been interpreted by the High Court to be a very narrow provision for a criminal offence. More broadly, we are very focused on electoral literacy. We had a significant increase in our approach to electoral literacy at the 2025 federal election. That includes a focus on media literacy, digital literacy and literacy around artificial intelligence—we had dedicated comms and webpages with respect to that. That will be an area of ongoing focus for us, and that requires a sustained effort from us to help give the voter the resources and the tools and the confidence to navigate the information environment in which elections are now delivered. Then, of course, one of the cornerstones to us protecting the integrity of elections is the Electoral Integrity Assurance Taskforce, where we bring together multiple agencies—law enforcement, intelligence, security and some other policy agencies—and there is a collective effort to very closely monitor the environment in which elections are being delivered, and share resources, information and intelligence to deeply understand the environment and to put strategies in place to try and manage the election through that challenging information environment. That's a broad brush, but we're happy to go into greater detail if you like. CHAIR: I think you mentioned this, but your submission says that the AEC is considering a requirement for deepfakes in electoral communication to be labelled. In your view, where should the onus for this requirement sit? How would that work in practice? Mr Pope : For some context around that, the electoral act delegates the authority to me to produce a determination as to how communication ought to be authorised. The current determination is now five years old, nearly six years old, so it's time for us to update that because the environment has changed dramatically and technology has changed dramatically. We have started that consultation process. We've circulated a discussion paper to political parties, members, senators and significant third parties, and it's also on our website. The first round of consultation closed on Monday of this week. It's in that consultation paper that we proposed that one option that we could consider, subject to that consultation, is that deepfakes in political communication ought to be labelled in a simple way that just says, 'This content contains a deepfake.' The onus for that would sit on the person that's authorising the communication. CHAIR: Specifically in that authorising period of the election period? Mr Pope : Yes. There has been no decision made. We've got further consultation and legal advice and processes to go through. But that is one option, where it might just be a few additional words or some other form of label that advises the voter that that communication contains a deepfake. CHAIR: Is it just on deepfakes? What's the extent to which there should be AI labelling in general, particularly for political communications? Mr Pope : I think there are some challenges in that. The challenge with that is that, if AI is not already ubiquitous, it's not far away. If you're labelling everything with AI, then, at the next election or not too long thereafter, every bit of political communication could have an AI label on it, and you've got to wonder about the effectiveness of that. CHAIR: What are you, at the moment, in terms of AI amplifying and scaling or giving nefarious actors greater capacity than they've had before, looking at in how we are combating potential foreign interference in our elections? Mr Pope : I'd probably say at the outset that artificial intelligence best summarises the capability that amplifies and accelerates existing threats to electoral integrity; that would include mis and disinformation, influence operations, foreign interference and even elements of physical security threats through harassment, impacting social cohesion and other things. That's our perspective. I don't think this is in and of itself a new particular threat that needs to be dealt with on its own; it is a new threat that amplifies and accelerates existing threats to electoral integrity. Foreign interference would be included in that. The lead agency on that is Home Affairs. Home Affairs are a member of the Electoral Integrity Assurance Taskforce, so it's through that construct that we'd be engaging with them on anything to do with that. Thankfully, so far our elections have been absent any form of foreign interference. CHAIR: But that's where that process would work, through that— Mr Pope : That's right. CHAIR: Thank you. Senator Cadell. Senator CADELL: You mentioned deepfakes. We'll go back to the very famous reel—we won't go hypothetical—of Senator Pocock as a 'Supergreen', with a Greens shirt on underneath his clothes. That was generated not by AI but by artwork. Would that be categorised as a deepfake when it's not used as AI, under the current standard? Mr Pope : As Senator Pocock knows, there's been technology around for a little while that superimposes images— Senator CADELL: It was Canberra. It was hand-done. It was artwork. Mr Pope : like Photoshop and other things. We haven't settled on a definition of what a deepfake would be exactly. I note that some other jurisdictions are using a definition that is effectively along the lines of 'the use of your image or voice without your consent'. In that sort of scenario, that would most likely be the case, yes. Senator CADELL: On this consultation you're speaking about, I know that in the Farrer by-election there were livestream issues around authorisations, because it's very hard to put authorisations on a livestream; it is a very difficult thing to do. But they are so genuine. It's much easier for me to generate a deepfake of me doing something and to put a full-time authorisation on that than it is to do a live video. How do we encourage real information and real things without penalising authorisations? Mr Pope : That's one of the things we're grappling with. As I said, the way in which people use technology has changed dramatically. We are grappling with these sorts of scenarios in terms of what the new authorisation determination might deal with. The determination is narrow in its scope—so there will potentially be things that people want to be dealt with in the determination that we just don't have the legislative authority to deal with; that would be a matter for legislative reform, for parliament to consider. We're working our way through that. I don't think we'll have that finalised until well into next year, as we go through another round of consultation with a proposed determination and so on and so forth. Senator CADELL: You mentioned the ability to scale bad actors, bad things, foreign influence and poor things here. My concern is the resourcing in this new world to deal with it. I've been a campaigner and I know those last-minute emotional decisions are important; I know the mornings of are important. Now that any bad actor could scale misinformation, disinformation or deepfakes on the morning of, what resources do you need to be able to turn around quick enough and get rid of them? We've all done that morning thing where we've seen a sign we don't like, phoned you and had someone come and look at it an hour later, and it gets taken down. What do you do when there are 4,000 posts? I know Volodymyr Zelenskyy and his team are doing a great job of taking down Russian data centres at the moment, but they're not going to get all of them by our election. Mr Pope : It's clearly a critical challenge for us. The election is a huge undertaking. It is massive, as you would appreciate. They're getting bigger and bigger with every election, naturally. So widespread regulation of the communications environment is just a mammoth task. We've started this conversation with the Joint Standing Committee on Electoral Matters. We've put in a submission with respect to the current legislation around authorisations. I note that they've made a recommendation in their interim report that the parliament ought to consider some legislative reform. Right now, to your point, we have two options under our legislative regime. One is to pick up the phone to you and say: 'You've breached the authorisation determination. Can you please take it down?' The other is, if it's a breach of legislation, platforms may take it down because it's a breach of their rules. That has been increasingly less likely as time has gone on and they've maybe softened the interpretation of their rules. We're experiencing less take-up of that. The third option is for me to instruct AGS to go and get an injunction. I think the quickest time in which we could do that is probably about six or seven hours. A lot can happen at a polling place in six or seven hours. So we would need more powers to be able to deal with the contemporary information environment at the pace at which it moves in a way that gives us a chance of trying to do something. Senator CADELL: It is a scale. It's a hyperscaling of this. You can create 4,000 accounts and run stuff that no way you could keep up with. Mr Pope : That's right. It's further complicated, as you would appreciate, by the fact that enforcing Australian legislation in an international context on people outside of our jurisdiction in a timely way is incredibly difficult. So there are some real challenges that I think the standing committee and parliament need to grapple with. Senator CADELL: Okay. Thanks. CHAIR: Senator Pocock. Senator DAVID POCOCK: Thank you very much for your time, and thank you for the great work that you do for our country. The former commissioner warned a Senate inquiry looking at AI a few years ago now that AI could pose risks to democracy itself. He said: … the AEC does not possess the legislative tools or internal technical capability to deter, detect or then adequately deal with false AI-generated content concerning the election process, such as content that covers where to vote, how to cast a formal vote and why the voting process may not be secure and trustworthy. Has anything changed in that time? Do you now have the tools and technical capability within the AEC to do that? Mr Pope : No. This is one of the challenges with expanding our regulatory responsibility in this space. There's the scale issue that Senator Cadell spoke about but also the technical capability. We as a regulator are certainly maturing, and that's largely as a consequence of the funding and disclosure reform regime that we're currently implementing. So we as a regulator are growing. We've got some increased resources, but those are purely to cope with funding and disclosure reform. So any expansion of regulation in this space will require some additional investment and sustained investment. That includes both people and expertise but also technology to assist us in trying to detect this. Senator DAVID POCOCK: You mentioned earlier you've been consulting on the possibility of labelling deepfakes. Who is the AEC consulting with? What does that look like? Mr Pope : We sent a discussion paper to all registered political parties, all registered significant third parties and other political participants. We put it on our website. We put out a media release. I thought we'd sent it to members and senators as well. I'm told it was not individually. That's been out for about a month. As I said, comments closed on Monday. What we will now do is work on understanding what the feedback is. We'll work then on a draft determination that takes account of that consultation and then that draft determination will go out for another round of broad consultation for further opportunity for comment. Senator DAVID POCOCK: This may be veering into opinion territory, but my sense in talking to psychologists and others is that it's one thing to label deepfakes. But as a human, when you see someone say something, you still believe it. Even if there's a label that says 'deepfake', should we be looking to potentially just ban deepfakes? I personally don't see the benefit of deepfakes in our elections, given the risks that they pose. Does the AEC have a view in terms of how you draw that that line when it comes to this technology? Mr Pope : Not really, and I'm not in a position to comment on your perspective there and maybe what the research is suggesting. I don't know how we could ban AI and put that genie back in the bottle. I can't get my head around how that could be particularly regulated at election time. Senator DAVID POCOCK: If you're deepfaking another politician, it seems pretty clear to me that, if you don't have their consent, that should be. Outside of elections, I think it should be— Mr Pope : That would be a matter for parliament. That would require legislative reform if you want to do that. South Australia has passed that legislation. New South Wales has passed that legislation, but it only takes effect from the issue of the writ for the election period, so it's a relatively narrow period. But I think the scale of trying to detect and regulate that would be incredibly challenging. Senator DAVID POCOCK: Finally, the AEC really pushed back against administering truth-in-political-advertising powers before the last election. Do you believe some other agencies should have these powers to deal with AI? What's the view there? Mr Pope : Before I answer that question, on the back of your last question, it was remiss of me not to mention that if parliament were to consider such legislative reform around banning deepfakes or artificial intelligence then it would need to also consider the implied freedom of political communication. That's a matter that parliament and others would need to work their way through. There is that. That's always in the background when we're talking about communication. With respect to truth in political advertising, our position hasn't changed over the years. It's not that we're critical of the concept or the intent; it's just not for the AEC to administer. We feel that that would put our impartiality at significant risk. Our impartiality is incredibly important. Back to my opening comments about trust and confidence in the AEC, in the electoral process and in the outcome of the electoral process—that flows through to the trust and confidence in you, parliament, and the institution. I think there is a risk that, the more we're asked to regulate and to make judgement, there is a risk around that impartiality being challenged. That is something that I would ask parliament to consider very deeply when it's considering giving us further regulatory responsibilities or whether it best sits with other agencies that don't need to grapple with that very fine-tuned impartiality. Senator DAVID POCOCK: Thank you. CHAIR: Ms Mascarenhas. Ms MASCARENHAS: First of all, thank you for the work that you do. I'm going to say that I think democracy is cool, and I'd argue that Australia has the best democracy in the world. From my perspective, I'm going to say there are five reasons: compulsory voting, preferential voting, Saturday voting, pre-poll and postal voting, and of course democracy sausage and cake stalls, which I'm going to say make a cash injection for many of our schools across the nation. But there's this other thing as we think about risks, which I'm going to say is physical ballot papers. If we think of the way that AI can potentially manipulate things, physical ballot papers are fashionable again. The thing that I am wondering, because of our compulsory voting system—what it means is that AI has an ability to massively influence the entire population of the nation, which I see as a significant risk. For Australia to continue to be the best democracy in the world, what do you think are the key things that we need to do with in relation to artificial intelligence and possible interference in elections, and does the AEC have the capacity to be able to do that? Mr Pope : You're right. As I said earlier, AI can amplify and accelerate existing threats and mis- and disinformation and influence operations, which are existing threats that could be amplified through artificial intelligence. I think the point that I'd make, which can sometimes be easily overlooked in some of the general discussion—it has always been the role of the voter to form their views, seek resources, seek information, challenge information, form their opinions and then cast their secret ballot at the ballot box. AI doesn't change that, and nothing should or ever will change that, but it's another challenge for the voter to navigate their way through. I think the best approach—there may be some regulatory responses that parliament might want to take that provide some assistance to the voter. I think labelling deepfakes—so that calls out what it is—is one of those that could provide assistance to the voter. I think the greatest investment that Australia can make is in civic education and electoral literacy, which need to be uplifted and sustained and widespread. It really needs to be seen as a core element of Australia's societal fabric, because, going back to your point, we need to preserve this winning formula that we have around our democracy in Australia, and it is under threat, and democracy is backsliding internationally. I think sustained investment in that area is absolutely critical. There are lots of good initiatives going on. There are discussions around being more coordinated. It's got to start in schools, but it actually needs to also go all the way through adulthood, and we are doing what we can with what we've got, but we could always do more if we had more support and resources to be able to do more. Ms MASCARENHAS: How effective do you think the civics engagement component is, and, to basically insulate Australia from that, how much would you need? Mr Pope : From how— Ms MASCARENHAS: To educate the Australian public, and that could be a question on notice. Mr Pope : I may well want to supplement my answer with more detail on notice. From my understanding of the environment, the education departments across the country are trying hard, but it's a contested curriculum. I think there's a discussion potentially to be had about curriculum priorities. That's not for me but for others. About consistency across our federated education system is a question that I have. The challenge that we have is that most people really engage with our education material just in time, just before the election, just before they're about to walk into the polling place or just as they're trying to understand how to apply for a postal vote. So it's about actually energising all Australians to really love and preserve our democracy and invest in it with civic education and seeing that electoral literacy and civic education and really understanding preferential voting and electoral processes as a core asset and being invested in that. That's going to require a lot of work, a lot of effort and a lot of coordination across many, many sectors and departments from school and from citizenship and all the way through people's voting lives. Ms MASCARENHAS: I went to a party which included children on an election day. It included a vote on who you support more, KPop Demon Hunters, Bluey and other groups as well. You got to vote below the line. What was cool is that we were speaking the language of children and using it in things they actually cared about. That helps with translation and helping see preferential voting happening in real time over issues or things that they care about. My final question is: we've seen social media polarise our communities and, from an AI perspective, it kind of averages information, so there's a possibility that AI could play a role in bringing communities together to have less polarisation. I'm unsure if you've got any thoughts or comments on that. If not, that's all good. Mr Pope : I don't think I'm in a position to comment on that, but, if you don't mind, I might just add, supplementary to my last answer, that the one thing I probably overlooked is that when I talk about electoral literacy, I'm also talking about media literacy, digital literacy and literacy around artificial intelligence. It's all critically important. There are some good efforts going in in different sectors around that. I also want to emphasise that one thing I've been reflecting on since the last election is that, with the way the information environment is changing and the resources that voters need to navigate it, the role of the media is becoming even more important as that other element of democracy that can actually help explain, highlight, emphasise or call out. They have a critical role in everything that I've just spoken about. That's an area that we'll be continuing to look at strengthening our very good and collaborative relationship with into the future. Ms MASCARENHAS: You've made me realise the other question that I had which was: we've been talking about sovereign compute and making sure that we've got compute that is for the good of Australia. Do you have an opinion on whether we need a certain quantity of compute available to fight the AI bots that could come out during an election? We would need an estimate of how much compute, and we would need to think about how we do that. What's fascinating is the interaction of election spend and how we work out what that compute quantity is. Mr Pope : You've gone well beyond my level of knowledge and expertise. What I would say is that, in our context, humans alone can't fight machines successfully. There's going to have to be some form of advance in technological capability here that's going to have to help us defend democracy and to combat those bots. I hold significant concerns about the impact of artificial intelligence on cybersecurity. For us, that is already an existing threat vector to electoral integrity as we protect the national roll system and our electoral management system and other key systems. We are seeing increased activity that's been going on for quite some time; I expect that's going to continue. We are doing our very best to protect our systems with some of the very best humans who are working day and night in our cybersecurity team and our infrastructure team. They are patching software left, right and centre and doing a fantastic job, but humans can only do so much. There is, hopefully, a positive role there for technology to help defend against technology and help protect our systems from the significantly growing cybersecurity threat. The resources required for that at the moment are quite considerable. CHAIR: Ms Chaney. Ms CHANEY: Do you see any legitimate use of deepfakes in political ads from a freedom-of-political-communication perspective? Is there a proactive argument for any use of deepfakes? Mr Pope : It's not really for me to answer that, I don't think. I don't have any real comment at all on the contents of political advertising and communication. That's for the parties and for others. Some might say there's satire and humour and other things that could be used. But I don't have a role or a comment or opinion on that. Ms CHANEY: Going back to the nightmare scenario where, the day before the election, there's a deepfake video of a political leader that actually changes the outcome of the election, it sounded like the options are, if you can find who made the video, to tell them that they've breached something; or to try and get an injunction, which could take six or seven hours and it may have gone nuts by then. I'd love to hear more about that third option, which was asking platforms to take it down, basically. Are these your only tools, and, if so, how do we strengthen that ability to get platforms to take it down—because that seems like the only really viable one there. Is there anything else that you need in your armoury to deal with that situation—because it doesn't seem like an unlikely scenario. Mr Pope : No. I think that's becoming increasingly possible, regrettably. We saw a deepfake in the Irish presidential elections that was somewhat along those lines, where it was a deepfake of, I think, the main candidate announcing her withdrawal from the election and it cut to a newsreader who was then broadcasting it. It was actually a high-quality deepfake. Those sorts of things are concerning. Deepfakes about even me saying something on election night about how an electoral process may well have changed which creates all sorts of confusion or turns people away—there are a whole raft of concerning possibilities. Before I answer your question about what else we would require, right now, yes, as you've outlined and as I outlined earlier, we've got some limited capability in terms of the regulatory levers we can pull. In that particular scenario, we would be doing significant media debunking and getting the message out about our view that that was a deepfake—that people are not to pay attention to it. That goes back to the comment I made earlier about the role of the media being so critical. At the moment, some platform rules—and I can't say all—are that, if there is a breach of your legislation, then they will undertake activity to rectify that breach. So a breach of authorisation requirements is a breach of our legislation, and on occasions they will remove it. I'd like to think, in that particular scenario, they would understand the severity of it and they'd remove it pretty quickly. We have put a submission in to the joint standing committee, a dedicated one with respect to authorisations where we've spoken about maybe the need for us to be able to issue a notice to direct certain material to be removed and for it to be an offence if that's not complied with. That is something that parliament might want to think about, considering there are also better and quicker information-gathering powers that will allow us to progress an investigation. But, in your particular scenario, that's more likely to occur within the hours and days after it's been published rather than on the spot. But we feel that an ability to issue a form of direction that requires a certain activity to be undertaken—whether it be by a candidate or a party or a significant third party or a platform—and that having significant consequences associated with it for noncompliance would be something that would be helpful. Ms CHANEY: So at the moment you're just depending on whether the particular platform has rules that work. Do they not all have those rules? Mr Pope : I can't say that they all do. I can take that on notice and come back to you. But it's not just about whether they actually have that as a rule; it's also about whether they wish to apply it—and we get a variety in the application of that rule in those circumstances. Ms CHANEY: How do you see the Electoral Integrity Assurance Taskforce? How do you see its workload changing with the advent of AI? Do you think it's well-positioned to respond with current resources, given how you might see that changing? Mr Pope : Fortuitously, I have the chair of the Electoral Integrity Assurance Taskforce board to my left. As the Deputy Electoral Commissioner, that's one of the many roles that she performs. So I'm going to ask her to respond to that. Dr Gleeson : The Electoral Integrity Assurance Taskforce has been around since 2018. The construct is about coordinating partners from across the Commonwealth and cooperating within existing legislative remits to guard against threats to electoral integrity. The taskforce focuses primarily on mis- and disinformation, foreign interference, cybersecurity risks and physical security risks. The workload has increased as the environment has become more complex, and we expect that that will be the case at the next election. We've certainly been focused on how AI changes the threat landscape across those various threat vectors. The agencies that support us are deeply committed to the EIAT, and we do work throughout the electoral cycle in preparing, in running scenarios, in understanding the dynamism of the environment and how we might respond. We also test whether our existing remits are going to be able to guard sufficiently against the threats. There's no doubt it will be complex, where we're confident that the construct is robust. It's seen as world leading. We're engaged by other electoral management bodies around the world on the EIAT construct. That prework that we do to understand the environment also extends to working with the community. So we do a lot in-house at the AEC, as the Commissioner has mentioned, around electoral literacy and working with the community to educate them about the risks to electoral integrity and the role that they play in helping ensure a free and fair election and casting their vote in a way that reflects their choice. We do that work with EIAT partners through some of their community liaison resources, particularly with the Department of Home Affairs and others. Ms CHANEY: Is it mostly done through other agencies? I guess my question is about its resourcing, whether it has adequate resourcing, whether it needs it. Dr Gleeson : The EIAT is resourced by—it's in-kind support from relevant agencies. So it's not a funded construct. Ms CHANEY: Do you feel like it's well equipped for the next election? Dr Gleeson : It's very challenging. It's a complex environment. As the environment has become more complex, I don't think any of the relevant agencies would say that other work has dropped away in their relevant patches. But that's, of course, a matter for them. We do the very best we can, like in every other area of our work, with the resources that we have. Mr Pope : I might just go back one question, if you don't mind. I neglected to mention that, if, in your scenario about the deepfake, it had come from a foreign entity, then we'd also have some power to work with platforms about breaching our legislation because they're unable to authorise political communication in Australia. So there's another possibility there. Ms CHANEY: What about if it were a deepfake authorised by a political entity but is misleading and deceptive? If I put out something and put 'authorised by Kate Chaney' on the bottom of it, and it was a picture of the Prime Minister biting the head off a puppy, but I've authorised it—I can't think of an example off the top of my head, but something that could change the outcome of the election using deepfake technology—what's the recourse there? Mr Pope : That sounds like it's properly authorised and would be acceptable. Ms CHANEY: So it's only the authorisation. Mr Pope : It's only the authorisation requirements. Ms CHANEY: So, if it's misleading and deceptive—if it's a video of the Prime Minister appearing to give a bag of cash to someone dodgy—and I've authorised it, there's no recourse? Mr Pope : No. That's right. Under our legislation and under a number of High Court judgements—section 329 of our electoral act relates to misleading and deceptive publications. But it's very narrowly constructed and interpreted. It relates to misleading and deceptive publications that relate to the casting of a vote, and it applies, really, only at election time. So, if it were a deepfake of a senior person in government saying, 'The only way your ballot paper will count is if you complete it this way—'1' here, '2' there—and anything else won't count,' then that may well fall within that very narrow construct. But, then again, it would have to happen at the time of voting. If it happened now, it would probably even fall outside of that. I guess there's also another question there for parliament about whether they want to reflect on some of the High Court judgements and the current, narrow interpretation. But, again, up against the implied freedom of political communication, there's a balance that needs to be struck. Ms CHANEY: But, beyond defamation laws, there's actually nothing that could stop something very misleading and deceptive from going viral the day before which shows someone doing something that they never actually did and that could change the outcome of the election. Mr Pope : No. And that is where there's currently the authorisation. There'd be our debunking. There may well be a disinformation register. There'd be our engagement with media and widespread responses. There'd be a huge communications response that we would need to put in place. But that's right. Ms CHANEY: Yes. Okay. Thank you. CHAIR: Senator Dolega. Senator DOLEGA: Thank you for appearing with us tonight. It's almost like getting practice for estimates, I guess, for you. I want to follow on from Ms Mascarenhas's questions. The other morning, I read an article from the US that said a lot of people are using ChatGPT to decide whom should they should vote for in the midterm elections there. It got me thinking about how a lot of these more common models that people use, whether it's ChatGPT or Claude, are controlled by very, very rich and powerful people. Would training in Australia give us more control over what is being presented to people? I use the perspective of, say, if you could sponsor results. You might have an actor who says, 'I'm going to give you a bucket of money, and, when people ask about elections, I want you to give them this information about political party X.' If we had training here, would that give us more ability to regulate and control that so that it would be free from bias, or should AI not be able to give advice about elections at all? What do you think? Mr Pope : It's definitely an area of emerging concern. This comes back to our point about electoral literacy. I think a lot of the commentary on electoral literacy that we apply to social media platforms also applies to AI platforms and tools. You need to do your work and seek verification, validate information and consider a range of different resources before you form your own opinions. It's dangerous to be using social media platforms or AI tools as your only source of research when it comes to forming an opinion around voting. That's a critical message that we apply through our electoral literacy—that needs to be something that we continue, as we discussed earlier. In terms of other measures, it sounds like they would require, potentially, a parliamentary and legislative response about what platforms can and can't do and what they should and shouldn't do in terms of responses to information. I think there was a Demos report earlier this year that tested responses from generative AI tools in the lead-up to the Scottish election, and 34 per cent of responses from those tools were found to contain errors. That is concerning in terms of voters wanting to try and understand the process, where to vote, how to vote—how to cast a formal vote, I mean. We're making a significant investment at the moment into modernising our webpage so that AI tools actually pull factual information from our webpage and reflect it back to the voter, rather than pulling information from a variety of other resources, which may or may not be correct. I think that's part of this digital and AI literacy that we've spoken about. People need to understand that these tools aren't necessarily pulling factual information from the agency that's delivering the election or the experts who understand elections; they're pulling information from a whole range of resources, which may or may not be correct. But anything that regulates how those platforms present information, or how they don't present information at all, is going to have to be a matter for parliament to consider—and, again, in the construct of the implied freedom of political communication, potentially. Senator DOLEGA: I agree. Following on from your comment about social media, we see countless pages that will appear through algorithms that a lot of reports say are coming from overseas, trying to influence people and generate outrage. As communities and people get more and more used to using Claude, Gemini or whatever it may be, we're putting an awful lot of trust in these platforms to give people that information. Your insights are very, very curious and give us some things to consider, as I'm sure they will the electoral matters committee. They will be looking at this as well. Mr Pope : Whether it's AI tools or social media platforms, people need to be very mindful of the algorithm that they're feeding themselves. Senator DOLEGA: I wholeheartedly agree. Thank you, Chair. CHAIR: We'll finish there. Thank you very much. Through the secretariat, please respond in the next two weeks to anything that you've taken on notice. That concludes today's hearing. I thank you and our earlier witnesses, and I thank Hansard and Broadcasting for their assistance. Committee adjourned at 20:02