GARLICK, Ms Mia, Regional Director of Public Policy, Australia, Japan, Korea, New Zealand and Pacific Islands, Meta SEETO, Ms Cheryl, Head of Policy, Australia, New Zealand and Pacific Islands, Meta [10:01] CHAIR: Welcome. Although the committee does not require you to give evidence under oath, I should advise you that this hearing is a legal proceeding of the parliament and therefore has the same standing as proceedings of the respective houses. The giving of false or misleading evidence is a serious matter and may be regarded as a contempt of parliament. The evidence given today will be recorded by Hansard, and attracts parliamentary privilege, and is being broadcast live on the parliament's website. I now invite you to make a brief opening statement. Ms Garlick : Thank you, Chair. I apologise, in advance. I've got a slight head cold, so I apologise if I sniffle in parts of this. I mean no disrespect. I will also keep my distance so that I don't get anyone else, here, sick. First of all, thank you, Chair, Deputy Chair and committee members for the opportunity to appear before you today. I want to begin by acknowledging the Turrbal and Jagera people, the traditional owners of the land on which we meet today, and pay my respects to elders past, present and emerging, and also extend that respect to any First Nations policymakers we're appearing before this morning and any other First Nations people who are present here or watching online. We have followed the testimony that's been given to this committee, across its hearings, and we acknowledge the experiences and concerns that have been shared, particularly by First Nations people who've been directly impacted by online racism. Meta also understands the concerns about the role that digital platforms can play with respect to racism, hate and violence. We understand too that there are growing and specific concerns about racism, hate and violence directed at Aboriginal and Torres Strait Islander people, particularly in the light of the recent distressing events in Perth this year and in Melbourne last year. I thought it might be helpful to frame our discussion here today by giving you some background context on three areas of our work. First, is with respect to our content policies. Second, is our work with respect to the 2023 Voice referendum. Third, is our work with First Nations organisations and communities in Australia. With respect to our content policies, we are committed to ensuring that people can use our services safely and positively, and we recognise that we have a particular responsibility when harmful content targets communities that are already experiencing systemic disadvantage. Our content policies have been developed and evolved over many years, and they're informed by input from our community, external experts, civil society groups and regulators around the world, including here in Australia. We enforce these policies, at scale, using a combination of advanced technology and human review, and we publish our results transparently, each quarter, so that the public and policymakers can assess our progress. In our experience, there are a number of aspects of our policies that protect Indigenous communities from online harms. These include our hateful conduct policy that prohibits attacks on people based on protected characteristics, including race and ethnicity, and our bullying and harassment policy that's designed to protect people from targeted abuse, intimidation and degrading conduct, and, finally, our dangerous organisations policy that prohibits terrorist organisations and organised hate groups from using our platforms. We remove content that glorifies, supports or represents them. In Meta's latest transparency report for Q4 2025, for example, we disclosed that we'd removed 1.3 million pieces of content on Facebook and 1.3 million pieces of content on Instagram for violating our hateful conduct policy. We acknowledge, however, that these numbers reflect a global scale and may not always match the experience of different communities every day. That's something we're committed to improving on. With respect to the Voice referendum, we certainly recognise our responsibility during significant civic moments, and, in the lead up to the 2023 Voice referendum, we took a similar approach to content integrity as we do for a federal election. That means that we stood up a cross-functional group, five months before the vote, to identify and mitigate risks, and, when we became aware of hateful conduct targeting Indigenous Australians, including harmful stereotypes, we took action. We expanded our Australia-specific slurs list, we created specific policy guidance for content reviewers and we worked to remove a significant amount of violating content. We consulted with our Australian combating online hate advisory group, which includes First Nations representatives, and partnered with the youth mental health service ReachOut to support the social and emotional wellbeing of First Nations people throughout the referendum. We also provided a funding boost to our third-party fact checkers to increase their capacity and coordinated with the Australian Federal Police and the Australian Electoral Commission. Lastly, before opening up to questions, we also work to support First Nations communities in Australia both within our company and through external partnerships. Internally our teams work to raise awareness around key cultural moments, making sure that as a team we're recognising National Reconciliation Week, NAIDOC week and other significant dates. Our facilities team actively identifies and supports First Nations owned businesses not only featuring them within the office but also going on to make sure that some of them become our regular suppliers. Externally, in recent years, we've partnered with First Nations led organisations, including Indigitech and Awesome Black, to create First Nations augmented reality content for Instagram. We've supported Indigital, Australia's first Indigenous EdTech company, to create immersive tech training for First Nations communities and supported the Awesome Black creator hub for Indigenous artists in Sydney. I hope that's some helpful context and look forward to your questions. CHAIR: Thank you very much for that opening statement. We'll proceed to questions and begin with Ms Cook. Ms KARA COOK: Thanks for coming today. I have a lot of questions for you—I suspect we all do—but I might just start with how this all works in terms of Meta. You've talked about the moderators. You've talked about the cross-functional group that you created specifically in response to the Voice, and you've talked about the slurs list. It seems like there are a number of different mechanisms. We know that people can report things themselves, but can you talk us through, for example, what each of those tools does and how it does it. Can you give us some specific examples, and, if you have any Australian data on what sort of numbers you're picking up through those mechanisms, perhaps you've got it with you today or it's something that you can bring to us later—perhaps on a monthly or weekly basis. Ms Garlick : Sure. Maybe it's helpful to start by how we think about it conceptually. The bedrock, the starting point, is our policies. We work to have policies that we do need to iterate over the 20 years we've been developing our content governance and integrity systems. That's because what's happening in the world changes and technology changes. We publish those, and they're quite specific in terms of outlining where we do and do not draw the line in respect of those four areas of the policies. There are 24 policy areas, but there are four that we wanted to highlight for this committee because we think they're the most relevant to your work. Then there's how we enforce those policies. Some of that is through these operational teams that I mentioned, who will receive user reports via the platform and then assess that content and remove it or take action against an account if there's been a violation. There are some areas of our policies where there is automatic detection. That will then run, and sometimes it will be where there's a high probability that it's definitely violating—a fake account or a known piece of terrorist content or organised hate content—then we'll automatically apply the strike or the action. But, if it's unclear, it will be surfaced up for the human reviewers. That's the enforcement bucket. As part of that, there is the slurs list that I mentioned. We recognise that we need to have moderation and the application of policies in ways that are locally relevant, so we have a market specialist team that's responsible for Australia and we work together with them to make sure that they can apply those policies in ways that work for Australia. And then we have additional tools and additional partnerships and awareness campaigns, if that makes sense. Hopefully, that's helpful. Ms KARA COOK: And region-specific data? Ms Garlick : We don't have region-specific data at this time. I recognise I was giving you numbers from our global transparency report, but we don't break that out into region-specific data at this time. Ms KARA COOK: Would you have any specific data in relation to First Nations people in Australia? Ms Garlick : I don't think that we would necessarily, in the regular course of our business, have that data. I think, in relation to some of our election reports as part of our transparency reporting under the misinformation code, we have given data about hate speech in Australia or certain types of content removals, but they're not broken out depending on the different protected communities. Ms KARA COOK: On the type of— Ms Garlick : Yes, that's right. Ms KARA COOK: How does that all work when you then look at the algorithm and how content is amplified, particularly racist content? We just heard from the Human Rights Commission—I think I actually wrote it down—that racism is profitable. So, from a Meta perspective, how does your algorithm work in terms of interacting with those policies? Ms Garlick : I think it's really important to understand that, from our perspective, we're trying to provide services that people find valuable and that they want to use, and we're also trying to provide a business where advertisers want to advertise on our services. Neither of those two groups of people want to experience harmful content on our services. There have been a number of instances over the history of our platform where we've had advertisers undertake boycotts because of concerns around hate speech, so it is absolutely not in our business interest to be promoting or providing a service where people are experiencing harmful content. I can talk generally about the algorithm and the different algorithms that exist and how they operate, and then I can talk about some of the measures that we take to try to make sure that we're not exposing people to content through that. There are a range of different algorithms. The first is a ranking algorithm. If any of you have used any of our services—you might use the service for a number of years and build up a number of different contacts, but you might not want to see everything from all of your contacts over the years every time you log in. So there's a ranking algorithm that will rank the content by your connections to try to identify if you might find that useful or interesting. Then, increasingly, online services are connecting people to unconnected content. That's content that you haven't liked or followed that we think you might be interested in based on a range of different signals. And then there are the security and safety algorithms that are trying to down-rank harmful content and trying to make sure that low-value content isn't being served to people. In terms of trying to make sure that we're not promoting harmful content, obviously there's the work to enforce our content policies to try to make sure that this content isn't on our platform to begin with. And then there are the content distribution guidelines and the recommendation guidelines, where we have additional policies to try to outline what type of content will be down-ranked and what type of content we won't recommend to try to avoid it being a harmful experience. And then, where we don't get that right, we have a number of on-platform tools to provide transparency to people about why they're seeing content and also to send feedback to us if they don't want to see particular types of content. So we do try to create an ecosystem where we're trying to continue to provide valuable services. Ms KARA COOK: Can you tell us about your moderators. Do they undertake, for example, antiracism training? What's their training or what are their qualifications to make these determinations about what is or isn't racist? I know that you talk about hate, but do you have specific training for those moderators around, specifically, racism? Ms Garlick : I think the best way to think about it is that the policies really are what govern how the moderators will apply the policies. Obviously, everyone's trained to be the best that they can at their job, but, for example, with our policies that are set, we have people who are expert in a range of different areas. Some are former prosecutors or lawyers or have been working for civil society groups, and they will be responsible for key parts of the policies. Then, where we identify the need for a policy change, for example, or we're starting to see a trend and we think we need to adjust the instructions to the moderators, we'll work with different expert groups around the world—that will be academics, that will be different experts and that might also be local community organisations—to get their feedback on where and how we can make a policy change that addresses the concern. Once that policy is set, we then provide implementation instructions to the moderators. So, whilst, yes, they will absolutely be trained in a wide range of the areas that they cover, they really are trying to implement a policy in a consistent way. I think, in terms of trying to answer your question, the focus really needs to be on the process that we have around the policy development and updating work. Ms KARA COOK: So, specifically for Australia, there is no antiracism training for your moderators? Ms Garlick : I wouldn't say that we have specific antiracism training for Australia, but, within the hateful conduct policy area, bullying and harassment, and the dangerous organisations policy and within areas that are trying to prohibit graphic and violent content, how we work to protect people based on protected characteristics, including First Nations communities in Australia, is part of the considerations that will go into where those policies are set and how they are updated to try to protect people on our services. Ms KARA COOK: Are the moderators in Australia? Ms Garlick : No, they're not. Ms KARA COOK: None of the moderators? Ms Garlick : None of the moderators. Ms KARA COOK: And none of the moderators are Australian? Ms Garlick : Some of them are Australian. Ms KARA COOK: But located elsewhere? Ms Garlick : Yes. Ms KARA COOK: Are any First Nations people? Ms Garlick : Not that I'm aware. Just to give you some context about why we try to co-locate people—we find, from a support perspective but also from a consistency and training perspective, that it's better to have people sitting together so that, as new issues come up or there are questions, there's a more consistent standard. That's if we have people co-located together. That's why we've tended to try to place people in the same offices when they're making decisions on moderation. Ms KARA COOK: Do you have any advisory groups or individuals who are First Nations people? Ms Garlick : Yes. I mentioned the online hate advisory group. I'm not sure if you want to share more about those, Ms Seeto? Ms Seeto : We have an advisory group in Australia that is a combating online hate advisory group that we established around 2019/2020. That is a group that is comprised of a mix of individuals. Some of them are representing specific communities from across Australia, including First Nations communities, LGBTQI+, youth and Jewish and Muslim communities. We also have within that group a number of experts who are academics or researchers in areas including hate and preventing violent extremism. We meet with that group once or twice a year and really draw on their expertise and advice as we work to make sure that we're applying locally appropriate enforcement on our policies. Ms KARA COOK: Do you have a written definition of what is racist? Ms Garlick : I think, with our policies—this is why I was trying to allude to the four different areas—we've got a definition of what is hateful conduct. That is conduct that is targeted at people based on protected characteristics, including race and ethnicity. Ms KARA COOK: What definition do you use for race and ethnicity? Ms Garlick : All races and ethnicities, I think. I'm happy to come back to you with more details if I can elaborate on that. Senator THORPE: Is it Donald Trump's, or where does it come from? Ms Garlick : We've built it up over many years and we've tried to consult with a large group of people to try to make sure that we're understanding, for example, international human rights definitions and things like that. I can understand very much the focus on hateful conduct, but, in terms of some of the negative experiences that First Nations communities or different protected vulnerable communities can experience, we do find that a number of policies work to protect them. What we try to do is look at some of these international definitions or human rights definitions and try to make sure that we've encapsulated them across a range of different areas. CHAIR: I'm going to take Senator Thorpe's interjection as an opportunity to rotate the call and go to Senator Thorpe. Senator THORPE: Could we get on notice who that advisory group is and who the Aboriginal person is representing Aboriginal communities in this country, please. Ms Garlick : Yes, we can take that on notice. Senator THORPE: How does Meta contribute to systemic, structural and interpersonal racism against the First Peoples of this country? Ms Garlick : I think, from our perspective, we're trying to provide services that are valuable and useful to people. One of the pieces of feedback that we've heard from Indigenous communities is that often they find online services helpful because they provide them with an avenue for connection, community and representation that they maybe cannot find in other fora. That's why we've had a longstanding commitment to try to make sure we are understanding the issues and concerns and can try to tackle harm to try to prevent our services from further contributing— Senator THORPE: Sorry, everyone has trouble with this question. I want to know how your organisation is contributing to the harm against First Peoples in this country. I want you to say how it's harming us. I don't want to know about what you're trying to do and that you've got your one little blackfella on your advisory and you might listen to them twice a year. I want to know how you're contributing to the harm. That's what I want to know. Ms Garlick : Obviously, it's not the intention; we'll just take that as given. The feedback that we've had from Indigenous community groups is that there can be issues and challenges where there are comments in groups or where there is content that is shared that is triggering and harmful and perpetuates harmful stereotypes. That's why we have the policies that we do and that's why we have the connections that we do—to try to make sure that we can work with our teams and try to take action against that kind of conduct or content so that people are not experiencing that. Senator THORPE: How much does Meta profit from this harm? You can take it on notice and bring back your accounts. How much do you actually profit? It's very profitable to be a racist in this country today. Go to my Facebook and have a look at that, and you'll see threats every day. You'll see constant abuse in my life as a black, sovereign woman. So how much do you make from that? How much does Zuckerberg make from our misery, which we see on Facebook? Yes, people use it to communicate, but there's violent harm. How much do you profit from that? Ms Garlick : In terms of how Meta makes money, obviously that's through selling advertising. Advertising is not connected to the type of content that's there. But, I think as I mentioned in some of the responses to the earlier questions, it's actually against our business interests for there to be harmful and hurtful experiences for people on our services. We face numerous advertiser boycotts from this. Even if you have scepticism around our desire to provide useful services for people, it's also not in our business interests for people to be experiencing this on our services. That's why we invest a lot in trying to make sure that people don't have those experiences. I have seen some of the testimony, Senator, about some of the threats that you've received. We're very happy to work with you and your team. There are a number of additional— Senator THORPE: The AFP are after the one in the US that Meta won't release. You might want to release that to the AFP so we can progress the Nazis who put it up. That's one thing you could do. Can I just give an example. If someone does an AI generated thing, like a racist meme through AI generation, and then there's advertising that comes from that, does your policy shut that off because it's advertising under false pretences? Ms Garlick : It doesn't matter whether it's created by humans or AI. Our policies still apply regardless. So, if content is on the platform that violates our policies, then it should be removed. And if there is advertising around it that is targeted on the basis of the individual—because when advertisers run a campaign they say, 'I want to reach all people aged 25 to 35 who live in Darwin,' for example—that's the basis for the targeting. There are additional rules that apply to any attempts to use AI-generated images for advertising that I'd be happy to share with you as a follow-up. We do try to have some protections around that. Senator THORPE: Thank you. And it's good that you've turned up, because you're facing the music, basically. But you can say whatever you like; it's clearly not working. Given that it's not working, and the rise of racism, hate and violence against the world's oldest living culture on the planet—they tried to wipe us out here, through racism; when they rocked up on their boats they came to wipe us out, and they failed, so they're continuing to try to wipe us out—what are you going to do? Everything you've talked about doesn't work, so what else are you going to do? Ms Garlick : Well, I would like to think that a lot of the work around making sure that our content policies are up to date and that our enforcement is effective, and we have small things that won't seem significant in the broader scheme of things—for example, as you mentioned, on comments, we've got a new program that we're trying to run because we have heard feedback that a lot of Indigenous organisations are experiencing a lot of hate through comments. We've got new initiatives that we're trying to roll out to support people through that. Senator THORPE: What are they? Talk about those. Ms Garlick : There are different tools, like the Moderation Assist tool. Some of the testimony before this committee—I can't remember which organisation it was—said that a lot of the low-value or hateful comments were coming from accounts that didn't have a profile picture or had things that made them look low value. The Moderation Assist tool is something we've developed over the years that will automatically block those types of accounts from being able to comment on a page. We know that sometimes there are issues where, for example, if a particular thing happens in an AFL game there can be a pile-on on a particular player. We've developed the Limits tool, which will stop people who haven't previously followed that account from being able to connect with it. So there are a number of things we're trying to do to try to create these safe spaces for Indigenous representation and voices. But also, as there's increased use of large language models for content moderation, we're in the early stages of testing and iterating that and evaluating it for efficacy. And we've seen some of the testimony before this committee that suggested that sometimes we were getting it wrong when we were taking down content that was trying to bear witness. We're optimistic that the use of LLMs will allow a more nuanced and contextual decision that can then account for some of these very complex nuances so that Indigenous Australians can bear witness to some of their experiences without feeling like we're imposing policies that are disempowering. So I think there's a lot more space, and I think that's why ongoing dialogue and collaboration together can be helpful and hopefully a reason for optimism. Senator THORPE: Well, on that—'collaboration together'—you've got one Aboriginal person for the whole country who advises Meta twice a year. Do they live here? Ms Garlick : I think it's more than one, but—do you want to explain a bit more on the work for the Indigenous communities? Ms Seeto : We have the advisory group, and they are Australia based, and we can provide more details on notice about the group itself. But we also have done a number of other partnerships with Indigenous organisations, which Mia mentioned in her opening remarks. I'm happy to talk more to those. It's a broader group of organisations including Indigitech, Indigital and Awesome Black. We've consulted with those groups. Also, as an example, I think you questioned the way we have taken into account the advice that we receive from the advisory group. One that Mia alluded to, which I can elaborate on a little further, is a project that we're working on at the moment. It's a new one, in response to feedback that we have gotten from that combating online hate advisory group over, particularly, the last six to 12 months. The social media managers and content moderators of community organisations, including First Nations community organisations, do experience quite unique challenges in that they are often navigating and having to moderate and deal with content that may be hateful and offensive towards their own community and their own identity. Recognising that, we are partnering with one of our longstanding youth partners, Project Rockit, to set up a new program, which will work particularly to provide support, training resources and a place of community to come together and share experiences about navigating those roles and how to connect with and make the most out of some of those tools that Meta has built to give those organisations more control over the content that they're engaging with on their own platforms. Senator THORPE: It's really putting the onus back on the victim though, isn't it? Ms Seeto : It's intended to— Senator THORPE: All the perpetrators get to make money, and— Ms Seeto : It's intended to be in addition to the policies and the enforcement that we have in place, which Mia has responded to, and also really responding to feedback that we have gotten that it is important to provide additional tools and to try and empower and train these particular individuals. Senator THORPE: Is there any thinking around having Aboriginal people provide advice on a more regular basis? I'm sure there are bodies, in this country, of Aboriginal people that could do that, rather than a few businesses here and a couple of advisory people there, which would be very difficult as Aboriginal people when you're dealing with a—if you're gay and you've got a disability and you're black. We tick all the boxes. So to sit with an advisory body that has other vulnerable groups makes it even more difficult for us to get a say, sometimes. So is there any thinking around having a dedicated national body here that we can go to as consumers? Then that could provide the evidence and the feedback to Meta to have it rectified in a more timely manner and a more comprehensive manner. Is that something that you could do? Ms Garlick : We've certainly changed up how we've tried to engage with different Indigenous academics and organisations over the years, and we've tried to be mindful of what you've alluded to—that sometimes trying to ask for advice and assistance can add to the burden. We'll look at the recommendations of the committee with interest to see how we can continue to iterate— Senator THORPE: If it were a recommendation, would you support it? Would you reinforce it? Ms Garlick : It's hard to commit to things that we haven't seen. Certainly, you can take from our statement that we do take these issues very seriously and we're working to make sure that we're continuing to evolve and be responsive to the issues and concerns of Indigenous Australians. CHAIR: Thank you, Senator Thorpe. I find it pretty extraordinary that you don't have any of your moderators in Australia to fully appreciate the Australian experience of racism. I find that at odds with your comments that you take these matters really seriously. My question, I suppose, goes to thresholds for harm. In my experience and certainly the experience of others—sorry, I realise I was supposed to go to Mr Venning on the phone. I'll come to you next, Deputy Chair. The threshold seems too high for having content removed that is harmful, that is racist. When I've reported content, certainly when other people have reported content—there are some very recent examples, like the racist Aunty Lisa comedian, out of Victoria. Lots of people reported her content and it took a long time to get that racist content taken down. And I look at the now banned pages, like Action for Alice. That did a lot of harm—did a lot to entrench and really platform racist stereotypes about First Nations people in the Northern Territory. It took far too long for pages like that to be removed. I suppose I'm wondering how you can reconcile your comments about you taking these things seriously with the experience of people who are having an interface with your system who would say that those two things cannot be true at the same time. The experience of First Nations people online is that you are not taking our experiences and the harm that is done to our kids and our communities seriously, through how you act or fail to act on the harm that is done to us online. What is the threshold? And does it need to be lowered for racist content to be removed? Ms Garlick : There is the way the system is intended to work, in the sense that if people see harmful content that meets the threshold for hateful conduct under our policies— CHAIR: Ms Garlick, I think there would be people who would suggest that the system is intended to harm us. So I would be careful about how you frame that. I think there are definitely people who would suggest the system is working as it should, which is to harm First Nations people. Ms Garlick : When the teams become aware of content that violates our policies, we work to remove that as quickly as possible. It is challenging to give specifics without—if it's helpful, if you wanted to provide a specific post et cetera, we could provide details around that. In one of the examples that you've given, I think the videos kept getting reposted and so we were taking them down again and again. So there are issues around that. But I also think that some of this, hopefully, will improve over time as we get faster with LLMs and things like that. But— CHAIR: Sorry, what are LLMs? Ms Garlick : Large language models in artificial intelligence that can act faster. I apologise. CHAIR: So, in saying 'as we get better LLMs', are you intending to have more artificial intelligence make those decisions— Ms Garlick : Yes. CHAIR: rather than humans? Ms Garlick : Yes. CHAIR: Do you think that's problematic? How do LLMs actually understand and appreciate the experiences of First Nations people? Removing more humans from it, in terms of understanding the human impact, I'm not sure is the answer. Ms Garlick : The way in which a lot of the automated systems work is that you train them based on training data, and, unfortunately, we have removed a lot of hateful content on our services, and so we can use that as a basis to train up these models. I think that, when our content reviewers are looking at a particular piece of content, they're often looking at the content and maybe some account information. The reason we're optimistic about getting more nuance in through artificial intelligence models is that they can look more at the beginning and the end and look more at the overall context of what's happening and hopefully make the decisions faster and with a lower error rate. I very much appreciate—and I hope that my opening remarks also reflected this—and I can very much understand the difference between a lot of the experiences that people have when reporting content on our services. That is why we do try to have partnerships with different organisations, so that, if there is something that is happening within a community, where people don't feel like our systems are working as we intend them to be—which is to not expose people to hateful conduct—we can really jump on that and try to understand what's not working, what needs to be fixed, and to get better and improve over time. That's certainly the goal, and we've got a number of different strategies in place to do it. Also, I recognise your concern around where the threshold is drawn, and I do think sometimes that people can experience content which is offensive but doesn't meet the threshold of hateful conduct under our policies. Where there is a difference between local law in Australia and where our global policies are, we do work to restrict access to that content in Australia out of respect for local law—if, for example, we do get complaints from the Human Rights Commission et cetera. So there are a number of ways that we try to balance where we've drawn the line and people's experiences, in trying to improve but also to respect different community standards that exist in Australia. CHAIR: If a body like the Human Rights Commission or the Commonwealth were to change the laws about what the threshold should be that constitutes harm, then you wouldn't adjust your thresholds or your LLMs or your human people to make sure that that content is swiftly removed? Ms Garlick : We have global policies that we apply, and if there's a difference between local and global then we work to respect the local laws that have been applied by blocking that content from existing in Australia. Yes. CHAIR: I don't quite feel like I understand that. Ms KARA COOK: So are you saying that there is a higher threshold in Australia because of the laws here? Ms Garlick : No, I was trying to suggest to the chair—well, I think that you might say that some of the provisions are in the human rights body of legislation. We have restricted access to content for violating the Racial Discrimination Act for Australia. So I think the threshold there is lower. Ms KARA COOK: There are different rules that apply to Australia? Ms Garlick : That is one that I'm aware of that we have done. Yes. CHAIR: So what you said is correct. Ms Garlick : Yes. Sorry if I'm being confusing. CHAIR: If the legislation were changed in a way that impacted, you might say that global standards elsewhere are that the threshold is higher again, whereas it has been adjusted for Australia. Ms Garlick : Yes, correct. CHAIR: I think that some Facebook pages are very clever about how they incite hate and violence and racism towards First Nations people that probably wouldn't meet your threshold. But certainly the intent is there when you have a look at the comments. Tiles like, 'Do you think welcome to country should still happen?' or, 'Do you think we should have one flag in this country?' are all intended for one purpose, which is to punch down on First Nations people. Where does that fit? It doesn't necessarily meet the threshold for being offensive in your category, but it incites a wave of hate towards First Nations people and our belonging in this country. How do you deal with that? Ms Garlick : It's hard to talk in generalities without seeing the specifics, because it— CHAIR: The specific is that there is a tile that gets posted on Facebook—Facebook, I think, is a particular cesspit— Ms Garlick : Yes. CHAIR: with an Aboriginal flag waving, and it says, 'Do you think we should have one flag in this country?' That is a very specific example of things that appear on Facebook. Again, there may be a flag waving with a question on that tile that says, 'Do you think we should have welcomes to country?' These are very specific examples. These aren't generic examples; they are very specific examples, and they get pushed to me. I'm not following these pages. They get pushed to me. I imagine they get pushed to Senator Thorpe and other people on this committee. The recommender system pushes them to us—like I want to see this content! How do you deal with that? I don't imagine it meets the threshold for being racist or offensive or whatever, but the intention is very clear. Ms Garlick : Yes. I think you're right that there will be content on the services that is distasteful and offensive. Where I think the teams draw the line is where it's shared with a caption or if there are a large number of comments under it that are starting to move over to the threshold of targeting people based on race and ethnicity. Then it will cross the threshold of becoming violating. But I think that some of what you're explaining is some of the reason for the overall experience that people are having, versus where the policies are being enforced. I think that's why we need to continue to try to undertake efforts to work with different community groups to understand where these things are crossing the line. CHAIR: Do you only deal with it on a case-by-case basis—so it would take me reporting that particular post and suggesting that the comments on there are verging towards violating your community standards, so there's that instance—or do you look at the repeated behaviour of a particular page or the type of people that would get on there and comment or whatever it might be? What's the process? Ms Garlick : There are different actions that we can take. Obviously, there's removing a post that someone has, shared. If a person repeatedly shares content that violates our policies then we may take action against their account. For high, severe violations—so if it's something like a designated hate organisation, including some of the ones that we've designated in Australia—that will result in account removal. But if it is—I feel silly saying 'lesser violation'—bullying and harassment, then they might get a strike, and then it depends on how many strikes they accrue over time before there might be action on their account like suspension or account removal. CHAIR: What do you use to determine what constitutes a hate group? Is it the federal laws? Ms Garlick : No. CHAIR: So it's your own policies? Ms Garlick : In most instances that I'm aware of, particularly in Australia, we've acted before federal laws, and we look at different signals in terms of whether an organisation is indicating that their primary purpose is violence and hate. I'd be happy to share on a confidential basis—because we don't like to disclose it—what groups and individuals we've designated in Australia since around 2017 and 2019. And then we've updated that list since. CHAIR: Ms Cook, and then I'm going to go to Mr Venning. Ms KARA COOK: Just on that, so you don't publish the list of dangerous organisations or individuals publicly? Ms Garlick : Yes. Ms KARA COOK: The terrorist designation is based on a country's government's designations—is that what you're saying? You seem to have different categories. You've got dangerous organisations and individuals, and then you've got the terrorist designation. So that list is country or region specific? Ms Garlick : We'll certainly give regard to what is designated in a country. For example, I think it was in 2019 that we designated 200 white supremacy groups, and I think that was before the majority of them were on different national terrorist lists. That's just an area where we've got a system where we're looking at a range of different offline signals around intent to violence, and then we undertake a designation process through that. Ms KARA COOK:   Can you say how many people, individuals or organisations are on that list? Ms Garlick : I don't have that number to hand right now, but, again, if it's of interest to the committee, we can look at what we can share in camera. Ms KARA COOK: Just taking on what Senator Thorpe said around individuals that might be undertaking a coordinated campaign, either against individuals or against a class of people, are you then providing those details to the Australian Federal Police for prosecution? Ms Garlick : There are a couple of parts to your question. First of all, in terms of coordinated actions, we'll typically take action on that not necessarily based on the ideology. If someone's been designated as a dangerous organisation or a hate organisation then they're not allowed to have a presence on our service, and we'll remove them and take the actions that we can to stop them from using our services. But if there's some sort of coordinated action that's less than designated organisations then we have additional policies around coordinated inauthentic behaviour that may capture it as well. In terms of working with law enforcement, we have a dedicated law enforcement liaison team. We're responsive to different requests that they have and we meet with them on a regular basis to share information. Ms KARA COOK: Do you do that proactively? Ms Garlick : If there are emergency disclosures or threat to life, absolutely. Ms KARA COOK: So only a threat direct, not using a carriage service to menace or harass and not other breaches of federal legislation? Ms Garlick : Oh, we will if they're— Ms KARA COOK: You will proactively refer those people as individuals? Ms Garlick : We have processes for emergency disclosure, and then we have processes for reactive responses as well. Then we have teams that work to share information and understand the threat. Ms KARA COOK: Can you provide the number of people who have been proactively referred to the AFP because you've identified a breach of law? Can you give us that information as well? Ms Garlick : Mm-hm. CHAIR: Apologies for skipping the queue, Mr Venning. Ms KARA COOK: Sorry. CHAIR: I'll hand the call to you. Mr VENNING: Firstly, I just want to thank you for fronting up today. Unlike many other organisations, especially government departments, you've chosen to come and give evidence. I'm sorry I couldn't be there in person. Aboriginal and Torres Strait Islander communities are increasingly engaging with public debate online, and Meta has acknowledged the impact that coordinated inauthentic behaviour can have on public discourse. This can, of course, be extended into racism. Increasingly, networks of non-human or automated accounts are being used to artificially amplify content, manipulate engagement metrics and influence your virality algorithms, allowing those with sufficient resources to manufacture the appearance of public sentiment. As it relates to racism—and, I suppose, more so public debate—what specific measures is Meta taking to detect, remove and prevent these coordinated bot networks, and how do you assess whether your current systems are successfully preventing artificial amplification from distorting genuine public opinion? Ms Garlick : I think there are a couple of parts to the response here. First of all, we do have policies against inauthentic behaviour and accounts on the platform, and we publicly disclose how many fake accounts we remove. In Q4 of 2025, we removed 1.1 billion fake accounts proactively within minutes of account creation. Obviously, we have other systems that run on the platform to try to stop them from getting on there in the first place. But, that said, if there are accounts that slip through those initial controls, we have teams that are focused on identifying coordinated inauthentic behaviour, and we publish the results of the actions that we take to investigate those networks and undertake what we call strategic network disruptions in our adversarial threat report, which I think we publish once every six months and are due to publish the next one. So we can share some details about that work and those reports as a follow-up, if that's helpful. Mr VENNING: Yes, thank you. Do you have any specific data as it relates to Australia and maybe the percentage of accounts which you think are inauthentic today? Ms Garlick : I don't think we have Australia-specific numbers. I think, in our investor portal, we estimate maybe one per cent of accounts or something might be inauthentic, but sometimes those are not for nefarious purposes. Sometimes it's because people like to have an account for their pet, or they might have multiple accounts, even if our policies don't allow it for different parts of their lives. So, again, I'm happy to provide you with that precise number, if that's helpful. Mr VENNING: Thank you, great. Meta removes content that breaches its policies of hate speech, but many Australians are concerned about protecting freedom of expression, freedom of political expression. How does Meta determine where the line should be drawn between lawful but offensive speech and content that should be removed for promoting racial hatred? Ms Garlick : In the way in which we set our content policies, giving people voice is a sort of founding principle and then trying to make sure that people can express themselves safely is the other part of that. That's why we do work to try to look at where the threshold can be drawn for hateful conduct content to try to make sure that, where it's targeted at protected characteristics and it's using degrading or dehumanising language at either an individual or a group of individuals, we work to take action. We know that a lot of these things are being debated in the public sphere, so, as I mentioned when engaging with the chair earlier, we do recognise that there will be some conversations in the public debate and also on our services which will be distasteful and offensive which may not meet the threshold for being hateful conduct under our policies. Mr VENNING: Is there ever a situation where your policies, your community standards, differ from law? Here, in Australia, we've recently strengthened laws as they relate to hate crimes and hate symbols. How do you ensure your content moderation aligns with Australian law, where it may or may not differ from global community standards? Ms Garlick : Typically, when we're looking at a piece of content, we'll look at the application of our global policies as an initial matter, but, where we have reason to believe that it might violate local law, we'll work with our legal teams and outside counsel, or, if we've got a complaint from a regulator that suggests that it violates local law, then we take action on it to make sure it's not available in Australia out of respect for local law. Mr VENNING: Okay, great. The rest of my questions have been answered, so I'll hand back to the chair. CHAIR: Mr Smith. Mr MATT SMITH: It appears that the policies that are in place are working in direct competition with the algorithm that exists to create echo chambers. So, if I want to see a certain type of content, I will get that certain type of content regardless of how hateful that is. The use of LLMs and AI moving into the future is fine. I mean, they are just tools and they're only going to work as well as the instructions that they're given. The word 'primate', for example, is a very loaded term, a term that if used at Aboriginal and Torres Strait Islander people is racist. It has been said by your moderators not to breach community standards. If that's the case, how can we have any faith in the rest of it? Ms Garlick : I'm going to have to take that one on notice, because I'm not across the details of that instruction being given to our community moderators. I'm happy to come back to you with details on that. CHAIR: Do you have a fact-checking ability or capacity that you use? Ms Seeto : Yes, we do. We work with two third-party fact-checking partners in Australia. We work with Agence France-Presse and Australian Associated Press. They're our two fact checkers. They work independently. They are accredited by an independent body, the International Fact-Checking Network. They look at content that has reached a certain degree of virality and make a decision as to whether there may be a claim or a piece of content—a post—that they then choose to investigate and fact-check. They conduct their investigation and make an assessment. Depending on that assessment, we might take actions such as labelling a piece of content that has been assessed as either false or potentially misleading. We also can reduce the distribution of that content, depending on the assessment that the fact checkers have made. CHAIR: So you wouldn't remove it. You would label it or reduce its visibility. Ms Seeto : Under our misinformation policy, we remove certain types of content. It particularly focuses on misinformation that is potentially leading to voter interference or to imminent physical harm. That's under the 'remove' component. Then we also have 'reduce' and 'inform'—a three-pronged approach to misinformation policy. Where it is not meeting the policy threshold for removal, if it then is fact-checked by our fact-checking partners, we may also then reduce the distribution and also issue labels. There's an overlay that sits over that post to inform users that it has been fact-checked. CHAIR: Who does the fact checking? Are First Nations people involved in the fact checking? Ms Seeto : We have two the two fact checkers. I am not aware of the breakdown of their staff as to whether they have First Nations representatives on their fact-checking team. CHAIR: Do you think that's a gap? I feel like, if you've come to an inquiry about racism and we know there's lots of misinformation that gets shared about First Nations people online, and you can't tell me if there are any Aboriginal people in those fact-checking organisations—are they here? Are they Australia based? Ms Seeto : The AAP are Australia based. I think AFP might have some Australian based and also some overseas. What I do know is that, in the lead-up to the voice referendum, at least one of those fact checkers intentionally did hire a fact checker, I believe, who was First Nations to provide that additional capacity at that particular time. Senator THORPE: Who was that? CHAIR: The fact checker? Ms Seeto : The actual fact checker? I don't know. Senator THORPE: Will you take it on notice? Ms Seeto : We can try to find out, yes. CHAIR: It would be concerning to me if you had this capacity to do fact checking but there were no Aboriginal people involved in that fact checking. It'd be great to know from you—you said you don't necessarily know if those two organisations that do the fact checking actually have First Nations people employed and in what capacity they are employed—what they are actually doing online. Ms Seeto : Potentially another question to ask is, as part of their investigations when they're conducting a fact check, that they do consult with a range of different experts of relevance to the particular posts that they are investigating. So we can maybe try to get a bit more information about the breadth of expertise that they draw on as part of their fact checking. CHAIR: Yes. There's absolutely the question of who they go out and engage with, but I'm also interested in the organisation's capacity itself to understand and what their knowledge is. It would be good to have from you in writing what happens when somebody reports a racist comment. Who reviews it? What is the average response time? What proportion of reports results in content removal or account action? You've also spoken about impressive numbers in terms of the content that you've removed. I'm also interested in what is reported versus what gets removed. I imagine there are lots of things that get reported that wouldn't violate your community standards. And then I think you've talked about sharing data with Australian researchers or academics, including First Nations organisations, so harm can be measured independently. Is that right? Ms Garlick : It would depend on how you measure harm, but we do try to work with different researchers. CHAIR: As I understand, it's through Meta's Data for Good program. Ms Garlick : Data for Good is slightly different. I think there's the Meta Content Library that researchers can get access to to undertake research using data on the platform. CHAIR: How do you get access to that? I'm interested in knowing how many First Nations people and organisations have access to that data. Is it available in Australia? Ms Garlick : It is available in Australia, and I don't have the details to hand. CHAIR: I'm happy for you to take that on notice as well. It would be good if you could provide the committee details on how many researchers or organisations have access to the data as well. What rules or policies do you have about how that data is used? Essentially, my question goes to how you ensure that it's used for good and not to continue to perpetrate harm. Senator Thorpe. Senator THORPE: I want to go back to your business model, because controversy sells and clicks drive engagement and profits. Am I correct in saying that? Ms Garlick : No, I would challenge that. We publish systems cards that outline why particular content—like how the different ranking and recommendation algorithms work. Engagement and clicks are one component of it, but there's a wide range of other factors. But, in terms of how we actually make money, we make money by people seeing relevant ads and advertisers feeling like they get good ROI in terms of the ads that they pay us to show on the site. But that doesn't mean people have to be on the site for a particular amount of time or click on a number of things. If you go on and see one ad and that works for everyone, that's a great day. Senator THORPE: Okay. I'm just curious to know why Meta took down a video from the National Indigenous Times of an Aboriginal girl being brutalised by a racist cop in Perth. Why did that get taken down? Ms Garlick : I can talk at a high level, and then Ms Seeto can maybe talk about the specifics. I saw the testimony around this in particular, and I think this is one of the examples where some of that nuance that we're talking about makes it really hard for some of our enforcement systems to work effectively. In that instance, my understanding at a high level is that, if there is content that our systems identify to be violent, it might take it down. Obviously, in that instance, my understanding is that it was shared to bear witness and for awareness. I think that's one of the things that we've seen, particularly in relation to hateful conduct, and why we've tried to set thresholds where they are so that people can bear witness to some of this stuff. Did you have anything else to add on NIT? Ms Seeto : In terms of how we responded to that, someone in our team actually did see the posts that NIT made to draw attention to the fact that this had occurred and reached out to NIT to understand what had happened and then worked with our ops teams to rectify the situation. We understand that we did manage to proactively resolve it for NIT. Senator THORPE: It's just that that was taken down very quickly and Aunty Karen sniffing petrol was left up there for longer, which causes harm. It's a form of violence against our people. The only other question I have—and I think you may have covered it—is about the supports available for the violent racist attacks that you see on your platform. What's available to the innocent victims this racism is perpetrated on? Ms Garlick : To the people who are in the content or the people who are seeing that? Senator THORPE: The people who are seeing it and being violated—who get sick, who kill themselves because they've seen something on your platform. What do you do about that? How do you help? What do you do? What's the policy? Ms Garlick : In relation to violent and graphic content, as we've seen from this mistake that we made in the IT, we do have some automated systems that are trying to proactively find that content so that people are not being exposed to it, and we're taking action on it before they see it. On Facebook, we're proactively detecting and removing violent and incitement content 88.8 per cent of the time before people see it. On Instagram, it's 89.5 per cent. We're certainly aware that seeing harmful content can have an impact, so we try to get on the front foot to remove it before people can be exposed to it. There may be instances where there is harmful content on the platform, where it's being shared for awareness or for some other reason, and so it doesn't violate our policies. Then we put an interstitial on it so that people are aware that it is potentially violent—it doesn't autoplay; they don't automatically see it—so that we can try to make sure that people are only seeing it in an intentional way. Senator THORPE: You take it down. There's actually no duty of care for someone who watches that and has their life turned upside down as a result of seeing violent racism. Your action is to remove it or to put a warning on it and that's it. Is that it? There's no care afterwards for that person or the group of people. Ms Garlick : We do work with a range of safety and mental health partners in Australia as well. Senator THORPE: Who are they? Ms Garlick : I'll share an example. Leading up to the Voice referendum, as part of the consultation we did with groups around our approach and strategy to the referendum, we had some feedback that the First Nations community would be needing some wellbeing support at that time, and that that was an opportunity for Meta to provide some support. We partnered with one of our mental health partners, ReachOut, to develop a campaign in partnership with a First Nations creative agency. They engaged a number of First Nations youth creators to share messages of support to encourage mental health and wellbeing. They used communities generally across Australia at that time, but particularly First Nations youth. Hopefully, that's an example of some of the work that we do in partnership with mental health and wellbeing organisations. Senator THORPE: But once the damage is done, that's it. That person's left to their own devices, basically. You either remove it, put up a warning or get ready for a racist referendum that we should never have had. What about now? That's what I'm interested in. There are people dying watching what your organisation promotes. We see Zuckerberg in the courts in the US for people being addicted to the platform. There could be a potential lawsuit coming for the violent harm that your platform is perpetrating and platforming. Is there any other aftercare or is that it? Is that the end of your statement? Ms Garlick : This is an area where we're always happy to work with different partners to see if there are additional things that can be done. We do have a number of different tools and support structures on the platform. For example, if we are concerned, or if someone is concerned, that someone on the platform is suicidal, there are a number of different ways that we try to have on-platform interventions. When people log back into the platform, there'll be a range of different support tips and the ability to message friends or get in contact with specialised support. It does depend on the harm type or the experience that people are having. We have tried to work with different mental health and other organisations to provide interventions on the platform. I think it's an area that we're always open to continuing the conversation about. Senator THORPE: Thank you. CHAIR: Is it correct that Meta's policy exempts content deemed as political speech, humour or satire? Ms Garlick : I think it would depend on how it is expressed. We certainly do allow political speech, but a lot of that exemption has tended to move more into the misinformation space. I think you mentioned an example where there was a claim that it was satire but they still removed it for being hateful conduct. So it can depend. CHAIR: How do your human or automated moderators distinguish between legitimate political debate and weaponised dog-whistle racism targeting First Nations people? I'm wondering what the committee can have a look at. What guidance do you have or what are the things that you've got to decide when it falls into what category? Ms Garlick : I can see what we can provide to you. As a general rule, for political speech, we assume the public fora of either the parliament or the public debate are where accountability can happen. But I'm happy to see if there are more specific details we give within our policies and to our moderators around this precise issue. CHAIR: To my mind, it's also about ensuring consistency in how it's applied. In 2025, you had policy adjustments around hateful conduct online, which reduced proactive enforcement and expanded exemptions around public interest debate. What measurable change has occurred in the volume and prevalence of race based hate reported on Facebook and Instagram since that policy change? Ms Garlick : Just to give some context to that policy change, there was a shift in the hateful conduct standard. It was designed to marry up, if you will, how our policies were being applied with the public debate that was happening on TV and in other public fora. What we have done since then is made sure that we are keeping an eye on a metric that we have in our transparency report called prevalence. We have teams that undertake a sampling to identify, out of 10,000 views, how many are potentially hateful conduct. In the measurement of that, we've found that the prevalence of hateful conduct content has been steady since 2022. The teams that are working specifically on the hateful conduct standard are very focused on this, but so far they're not seeing that there's been a shift or an increase in prevalence. To the example that was given earlier, I do think that example where we'd taken something down even though someone was trying to bear witness would probably be less likely to happen now that we've made these policy changes. It is designed to allow more flexibility, because there are a lot of nuances around the ways in which impacted communities also try to bear witness to some of this content. CHAIR: It's interesting that you say that the prevalence of hateful conduct online has remained steady, when the experience of people is that it's increased. I'm interested then in what your measurements are and how you measure it, because the experience of people on the ground is that it's increasing and it's becoming more intense. It seems like there's something that's either out of whack with your data or out of whack with what we're hearing. Those two things aren't quite matching. So it'd be good to understand what you are actually measuring and tracking to come here and to say that prevalence has remained steady. To Senator Thorpe's earlier question, your algorithms absolutely optimise for engagement and there is research that demonstrates that high-conflict, racially provocative content often drives higher comment volume. Research demonstrates that. How does Meta ensure its recommendation engine does not actively amplify racist commentary targeting First Nations people? Ms Garlick : We have a number of different ways that we try to do that. Also I'm not sure, in looking at the testimony today, if I've seen that research. If you want to send that through, we can share some comments on that as well, just to make sure that we're addressing the concerns there. I think the first part of trying to make sure that the recommendation systems are not amplifying is to try to make sure that that content isn't on the platform. That's a lot of the work that I outlined at the start around having the policies, working to enforce the policies et cetera. Then there are additional policies and systems in terms of the content distribution guidelines and the recommendation guidelines where we're working to try to make sure that we're not recommending what we might call borderline content or other content that might be shocking or provocative. Then, where we get that wrong, there are also ways that people can send feedback in the system of, 'I'm not interested in this,' 'I don't want to see it from this person,' et cetera. So there are a range of different ways that we try to prevent that from happening. If people are having a bad experience on the platform, they won't keep using our services. CHAIR: I think there are lots of people who have no choice but to engage on online platforms. It's the only form of connecting with family or community or whatever it might be. So I don't know that it's as simple as people withdrawing. I think there are a bunch of people who have made decisions to withdraw from the platforms because it makes them feel unwell. I think I know the answer to this question, but your response might be interesting. Is Meta generating advertising revenue from engagement generated with comment sections or posts that contain unmoderated racial abuse targeting Indigenous people? Ms Garlick : Advertising is based on targeting people based on demographics and interest areas. It's not directed towards comments. I guess, philosophically, it's hard to break out engagement with comments versus engagement with the rest of the platform, but ad campaigns are not based on how long someone's on the platform or what they're engaging with. The ads system is very separate to the organic system as well. So, from our perspective, they're two different systems. But, having said that, advertisers are not going to want their ads appearing next to content that is damaging for their brand. So not only are we working to try to be true to the principles that underlie our content policies but also, from a commercial perspective, we have an interest to try to make sure that the broader ecosystem within which advertising is appearing is something that is going to continue to make advertisers want to advertise on our services. CHAIR: I might put some of my questions on notice because I note Ms Cook has got one final question and we are over time. Ms KARA COOK: I know you've watched some of the inquiry, so you understand some of the evidence that's been given about First Nations witnesses explicitly saying that they have suffered harm as a result of their interactions with different platforms online. Do you believe that Meta owes Australians a digital duty of care? Ms Garlick : I know that's being discussed right now. We've been broadly supportive of the principle of it. We haven't seen a lot of the details of it. But, in other jurisdictions where it has been enacted, we've been working to be constructively engaged and compliant with those laws. In Australia, I've mentioned our work to be respectful of the Racial Discrimination Act. We've also had 10 years of trying to work constructively to comply with the different online safety laws and rules that have come out. So, certainly, from a first-principles basis, we'll continue to work to respect the laws of Australia. Ms KARA COOK: So you would support a digital duty of care? Ms Garlick : On a principled basis, yes. But it's really hard to know the details at the moment. So far we have a survey and a two-page issues paper, so we don't have a lot of details. But, as a principle, yes. Ms KARA COOK: Is there anything that we haven't discussed today that Meta are doing to ensure that people are safe on your platform? Ms Garlick : It's actually a very good question. Is there anything else? Ms Seeto : I feel like we have had the opportunity to share a lot about our work. CHAIR: There's also an option to provide any additional information on notice if you walk away from this morning and think, 'Oh, I should have said that.' Ms Garlick : It's been fairly wide ranging and robust, and I imagine the questions on notice will be equally wide ranging and robust. We will engage thoughtfully as well. CHAIR: That concludes the time that we have with you. Thank you so much for showing up and being accountable to committees like this. It is really important, even though they are tough questions. I think we are both equally concerned with people's experiences online. So we appreciate you coming and answering the committee's questions in what is a really difficult conversation. You have been asked to provide questions on notice, so please forward those to the secretariat by 18 August. The secretariat will also follow up with you after this with the questions that you've taken. It's also an opportunity to provide any additional information to the committee that you think would be of interest. You'll be sent a copy of the transcript of your evidence from today and have an opportunity to request corrections to any errors that you see there. Thank you very much, again, for the time you've taken to appear before the committee. Proceedings suspended from 11:21 to 11:38 .