HUNTLEY, Mr Shane, Chief Technology Officer, Google Threat Intelligence Group, Google [by video link] LORD, Ms Rachel, Senior Manager, YouTube Government Affairs and Public Policy, Google [by video link] SAHEL, Mr Jean-Jacques, Senior Manager, Public Policy, Google [by video link] Committee met at 16 : 30 CHAIR ( Senator Whish-Wilson ): I now declare open this hearing of the Senate Select Committee on Information Integrity on Climate Change and Energy. I begin by acknowledging the traditional custodians of the land on which we meet, the Ngunnawal and Ngambri peoples of Ngunnawal country, and pay my respects to their elders past and present. I extend that respect to all Aboriginal and Torres Strait Islander people here today. The committee will be conducting today's hearing in person. These are public proceedings being videostreamed live by the parliament's website, and a Hansard transcript is being made. I remind witnesses that, in giving evidence to the committee, they are protected by parliamentary privilege. It is unlawful for anyone to threaten or disadvantage a witness on account of evidence given to a committee, and such action may be treated by the Senate as a contempt. It is also a contempt to give false or misleading evidence. Witnesses also have a right to request to be heard in camera. If a witness objects to answering a question, they should state the ground on which the objection is made and the committee will determine whether to insist on an answer, having regard to the ground which is claimed. If the committee determines to insist on an answer, a witness may request that the answer be given in camera. I remind senators of their obligations, under the Behaviour Code for Australian Parliamentarians, to treat witnesses with dignity, courtesy, fairness and respect. I remind everyone engaging in the public hearing, including witnesses and members of the public, that the Behaviour Standards for Commonwealth Parliamentary Workplaces require people to act respectfully, professionally and with integrity; encourage and value diverse perspectives; and recognise the importance of a free exchange of ideas. Bullying, harassment and discrimination will not be tolerated or condoned. After that preamble, I now welcome representatives of Google, via videoconference. I understand that information on parliamentary privilege and the protection of witnesses giving evidence to Senate committees has been provided to you. Do you have any comments to make on the capacity in which you appear? Ms Lord : I'm YouTube's government affairs and public policy lead here in Australia. CHAIR: Would you like to make a short opening statement? Then senators will ask some questions. Ms Lord : Thanks, Senator. We'd welcome that opportunity. Are you happy for me to start now? CHAIR: Yes, please. Ms Lord : Good afternoon, senators—Chair and members of the committee. We appreciate the opportunity to speak with you today. The open internet has enabled people to create, connect and distribute information like never before. It has exposed us to perspectives and experiences that were previously out of reach. It has enabled increased access to knowledge for everyone. But, like other communications channels, the open internet is vulnerable to the propagation of false and misleading information. Mis- and disinformation risk undermining Google's mission, and that is to organise the world's information and make it universally accessible and useful. That impacts the trust users place in our services as sources of high quality information that meets their needs. Our approach to managing mis- and disinformation across our products and services is based around a framework of three key strategies: firstly, making quality count; secondly, counteracting malicious actors; and, thirdly, giving users more context. We tailor our approach to each product, allowing us to apply these principles in a way that best serves the unique needs and opportunities of every individual service. We recognise that we have an important responsibility to our users and to the societies in which we operate to address the risk of harms associated with false and misleading information on our platforms. We've spent decades thinking about information integrity, and we continue to invest in adapting and refining our approach in response to a dynamic and evolving information environment. While there is no silver bullet, we are constantly striving to ensure that our response strikes the right balance between protection from potential harms and respect for the fundamental rights of our users. We look forward to discussing these issues with the committee. CHAIR: As there are no other statements, we'll go straight to questions. I was doing a bit of research prior to the meeting and looking at fact checking. The latest information I could find was from the Google Safety Centre, and it was an article titled Google's approach to fighting misinformation online. In that document, you state: In November 2022, Google and YouTube announced a $13.2 million grant … to support their network of 135 fact-checking organizations from 65 countries covering over 80 languages. Could you update us on that, please. That's obviously four years old now What other investments have you put into fact checking? Do you still have the same network? Have things changed since then? Mr Sahel : As you said, this was four years ago. What hasn't changed, if you will, is that we've had a multilayered strategic approach to ensuring the safety of our users across Google platforms. To do this, we need to employ a panoply of approaches and adapt to the evolving nature of the risks that users face online. As the issues, the concerns and the societal trends evolve, we evolve our approach. We continue to have a strategy that is based on both acceptable-use policies and community guidelines that state what the acceptable behaviours on our platforms are—and we enforce those proactively—but we combine that with methods to empower the users of our platforms with information, giving them context as well as broader initiatives, for instance around— CHAIR: Sorry—because we have a number of senators here and limited time, could I perhaps be more specific in that question. Do you still have third-party fact-checking arrangements in place? Mr Sahel : We've never had fact-checking arrangements as such. We've got ongoing relationships with a range of organisations, which have included fact-checking organisations, and we continue to work with experts such as those on a range of issues relating to deceptive practices. That's how that's evolved recently—for instance, to take into account emerging real concerns, such as the exploitation of AI through deepfakes. That's where, for instance, a lot of our focus is these days, to address those new emerging and worrisome risks. The point is that, although we continue to have the same approach of trying to safeguard our users across our platforms, we evolve as the risks evolve. CHAIR: Meta explained to us the other day that they have third-party fact-checking arrangements, as did TikTok, with groups like AAP FactCheck. So that is not the case for Google? Mr Sahel : That has never been the case for Google. We have support from various groups like that, but in different ways. CHAIR: What was the $13.2 million grant for, then? It says 'for a network of 135 fact-checking organisations' on your safety centre article. Mr Sahel : This was a general support. This was a grant rather than a contractual arrangement, so they never provided fact-checking services to us. This was to support broader efforts in society to help to counter deceptive practices. CHAIR: I also read online that you were considering an approach similar to that of X in terms of community notes. Is that what you have in place now? Ms Lord : I'm happy to take that. On YouTube, we did launch an experimental pilot in the US in 2024. We are collecting feedback on that from viewers, contributors and creators. But, at this point in time, we don't have any plans to expand that pilot to additional markets. CHAIR: That's US only, then? In layman's terms, then, how does it work in a place like Australia? Is there any fact checking, for example, for misinformation content on YouTube for Australians engaged on YouTube? Ms Lord : Yes, of course. Let me answer that for you for YouTube. As JJ mentioned, we do not operate, and have never operated, a fact-checking program on YouTube, similar to other platforms. We haven't used fact checkers to validate individual pieces of user generated content on YouTube, we have never shown fact checks on individual pieces of content, and we don't downrank or demote content based on fact checks either. As JJ mentioned, what we are focused on is actually elevating higher quality information and providing users with context for the information that they are seeing, and we use a number of levers and product features on YouTube to achieve this. CHAIR: I just got the Google policy off your website. With misinformation, for example, this comes under 'Prohibited and restricted content' and 'Fake and misleading content and behaviour'. On misinformation, you say: False, inaccurate, or deceptive information can cause significant harm to individuals, businesses, and society. For that reason, we don't allow users to post misinformation. You talk about 'deceptive or misleading health or medical information', 'harmful content that contains deceptive or misleading information about civic processes' and 'harmful content of newsworthy events or civil discourse that has been manipulated to deceive or mislead users'. So how do you stop content that is misinformation if you don't have fact checking? What's the simple process that you use to stop that misinformation being posted? Ms Lord : Thanks for your question, Senator. I just want to be clear that my answer will address YouTube, and we do have different content or policies across a suite of Google products. I'm happy to share with the committee YouTube's community guidelines, which are our content policies. We do have some misinformation policies there. As you would imagine, and as we've mentioned in our opening statement, misinformation is a complex challenge, and we do use multiple levers to respond to it. On YouTube, that will include setting some baseline guardrails around the types of content that we do and don't allow on our platform. But it's also using rankings and recommendations to elevate high-quality information and providing context to users using a few different product features. That could include things like information panels where we're providing information to our users at the top of search results or on video. We do have an information panel for climate change. That will provide a factual statement and direct users through to a UN landing page. We provide labels around publisher transparency, and then we have a few other features where we're really elevating high-quality content, especially when it comes to news and health information. CHAIR: Thank you. Just as a matter of interest, does Google differentiate between disinformation and misinformation? Do you, for example, monitor coordinated inauthentic behaviour and take down networks at that point of content moderation? Mr Huntley : I can take that one on the threat side. There's a range of different terminology, but we certainly have a range of policies, and we look at information under these different policies. My team, as part of Threat Analysis Group and also with trust and safety and our intelligence group, have been tracking coordinated information operations and campaigns along those lines for many years now, and we publish in our quarterly blog what we take down. We track these threat actors, and we certainly have policies against this coordinated inauthentic behaviour, which is also often known as disinformation. Certainly any such behaviour which is coordinated and trying to push a message which is deceptive in that way is against our platforms, and we've taken action. We are fairly public and transparent about the actions we have taken there. CHAIR: If I'm not giving follow-up questions about your answers it's because we don't have a lot of time. But we may do that in writing to you once we have gone back and reviewed the transcript. I note that in the safety centre information you provide about finding misinformation online you quote a specific time in history, between March 2020 and March 2021, with more than 50,000 fact checks surfacing on Google search and that you are heavily investing in initiatives that support media literacy and helping users find trustworthy sources. Is that fact checking, in terms of Google search, saying that you detected, fact-checked, 50,000 misinformation posts or is that something different? Could you also briefly tell us about your investment in media literacy? Mr Sahel : Perhaps I could address this, Rachel, if you'll allow me, since this is broader than YouTube or search. As I was saying, we've evolved things. What we would do nowadays, what you would find as a user, is that we check many more than 50,000 sources of content. What would happen is that we would give information to users, for instance, with tools such as what we have that's called 'About this image'. You see an image online and it comes up on Google search. What you can do is right click on it and it will give you background information on this image. It will tell you when this image was taken and whether it has been modified or created using AI, so you get a better feel for whether this is, for instance, a piece of news: 'Is this an image that has been just taken in the last few days or is it actually an image that's been recycled from five years ago that may be completely unrelated and be used in the wrong context?' That's how we try to provide users with a range of factually related information that helps them to take informed decisions vis a vis the content that they see today. This is certainly on thousands and thousands of pieces of content and in different ways. We really try to do it in an expansive manner. Ms Lord : I'm happy to jump in there with some additional information on media literacy programs that we've supported in Australia, if that would be of interest. I believe Google was one of the original founding funders for some of the digital literacy work being undertaken by the Alannah and Madeline Foundation. They partnered with YouTube on a suite of videos called Hit Pause, which is designed to support older teens and younger 20s users to spot misinformation and other behaviours online. They live on a YouTube channel. I'm happy to share those with the committee. We also provide funding support for Squiz Kids, who produce a really fabulous product designed to help school aged kids identify mis- and disinformation. The other thing I'd point out is that last year, I believe it was, we partnered with Life Ed to bring Be Internet Awesome to Australia, which is an online safety program for primary school aged children. A component of that is around spotting false information online. So there's a range of efforts across varying ages being undertaken and supported by Google and YouTube in Australia. CHAIR: Thank you. If you could give us some more information on notice or anything else you could add around your investments in that area, that would be great. I'll now go to the deputy chair, Senator Ananda-Rajah. Senator ANANDA-RAJAH: I want to follow on from the chair's questioning around inauthentic content, particularly content that is generated at scale. I turn to the ACMA report from last year, 2025. Figure 5—I don't know if you've got it in front of you—has a graph which refers to the number of inauthentic account enforcements in Australia that have been actioned, prevented, in the context of spam, scams and fraud. It's all a little bit vague. But from 2022, 2023 and 2024 you don't feature in this at all. It's LinkedIn, TikTok and Twitch but there is no YouTube, no representation from Google. Why is that? Ms Lord : I'm happy to take this question. I believe that ACMA report you're referencing is a summary of the transparency reports that signatories to the Australian Code of Practice on Disinformation and Misinformation provide. I could provide you with a copy of our transparency report, and you could take a look and come back to us if there are any data points that are included in the report that you're interested in exploring. Senator ANANDA-RAJAH: No. I'm interested in why the report that your representative body, DIGI, supplies to ACMA, which then comes to the government, doesn't have any information from Google regarding the number of inauthentic accounts that you've actually prevented. Why is it not in this report and you have a report? Ms Lord : With respect, Senator, the way the Australian Code of Practice on Disinformation and Misinformation works is that signatories, which include Google—we were a founding signatory of that report—are required to report annually on the efforts we take to meet those commitments under that code. We provide those reports to DIGI. They are published then on DIGI's website, and the ACMA provides its own report in response to all of those reports from the signatories. Senator ANANDA-RAJAH: There is evidence from ACMA that there is no inauthentic activity, according to this report, on Google's platforms. Is that even credible? Ms Lord : I'd have to take another look at the ACMA report. But we were, I note, like when I did review the ACMA report— Senator ANANDA-RAJAH: Do you not review the content? Ms Lord : Of the ACMA report? Senator ANANDA-RAJAH: The DIGI content that you supply to ACMA. Ms Lord : We produce it, so I read it multiple times. Senator ANANDA-RAJAH: You know the data. Is there no activity to remove inauthentic content on your platforms? Ms Lord : I'll let Shane speak to you about that. But I can say that in the ACMA report Google was called out as providing the highest number of data points under the Australian Code of Practice on Disinformation and Misinformation in our transparency reports. I think there's no shortage of data in our transparency report. I'd be happy to share that with you so you can read the primary materials. Mr Huntley : I'm not aware of the exact report, but, certainly, we do put an incredible investment into detecting and countering inauthentic behaviour on our platforms. As I mentioned earlier, even in the specific area of coordinated inauthentic behaviour from state sponsored actors, we publish quarterly. We have multiple teams of many thousands of people across the company—across our trust and safety efforts, across safe browsing and across our account security—taking action against large numbers of actors at large scales. In many cases, we should also point out, we have to interpret the scale of detection actually as a win when we take action against this stuff quickly. It is not necessarily how many they create; it's the effect they have. We've certainly seen that from some actors where they attempt to do a large amount of activity, but we immediately take action. They're unable to create accounts very quickly. We take them down immediately, so they try and create new accounts. They try and post a video; we take them down. So we actually are constantly fighting this battle worldwide, 24/7, with some of the best security and trust and safety people in the world taking action against these actors. I don't think anyone else is committing the same resources that we are towards fighting these different threats. Senator ANANDA-RAJAH: With regard to inauthentic content and coordinated inauthentic content, does Google profit from this sort of model in any way? Do you reward channels that publish large amounts of low-quality AI content, also known as slop or slopaganda? What is the benefit to Google for letting a little bit through? Mr Sahel : Through our various policies, we attempt to demonetise—that is, not pass on or allow advertising on—pages that would be carrying violative content. To the extent possible, indeed, there wouldn't be any monetisation of that violative content—certainly if we're aware of it. Mr Huntley : Much of our efforts are trying to surface the best content. Going all the way back, our mission is to make the world's information universally accessible. Certainly, it's in our interest to provide the best content to users—the most engaging and high-quality content. We find that that is of benefit to the user and also of benefit to us. We do not want to produce or put forward low quality— Senator ANANDA-RAJAH: Mr Huntley, I feel there's a disconnect between your worthy mission statement and reality. Are you familiar with the Reset.Tech report on synthetic propaganda that was published on 10 November 2025? Mr Huntley : I'm not sure I'm aware of the specific report. Senator ANANDA-RAJAH: It might be worth googling it. This report identified 150 YouTube channels. It was an exploration of politically driven misinformation and disinformation targeting, specifically, the UK Prime Minister, Keir Starmer. It found a lot of these channels were likely AI synthetic content, and there was a proliferation from one year to the next, from 2024 to 2025, a big uptick. They found that nearly 80 per cent of all channels were created in 2025. There were very clear similarities, in terms of descriptors, content, layout, branding across these channels, suggesting that they were actually coordinated and AI driven. They garnered 1.1 billion views. Essentially, the report says that YouTube, through its paid partnership program—in other words, your monetisation program—is stimulating the production of political content coming from anonymous creators. Are you familiar with this report? Ms Lord : I'll jump in here for YouTube. I'm not aware of the report. I would welcome some time to review it and understand it. I'm not in a position to comment on that, to be honest. Senator ANANDA-RAJAH: Okay. It would be great if you could take it on notice because this is clearly a problem on your platform. It's been raised by a civic group and it's not clear to me that you're doing enough to clean up AI generated slop on YouTube.    I'm worried that your paid partnership program, your monetisation program, is essentially skewed to eyeballs, so the more sensational or sensationalist and attention-grabbing the content—that is what is preferenced above quality. Am I right, or am I wrong? Ms Lord : You're asking a good question there. I think there are a number of facets here that go to respond to this question. We do set the bar for monetisation high through our YouTube Partner Program. There are thresholds for entry to that program, and all participants in the YouTube Partner Program are required to comply with a suite of policies that include community guidelines et cetera, our advertiser-friendly guidelines. I'm happy to share some more detail with you around that, but— Senator ANANDA-RAJAH: If you could focus in on the integrity aspects of those guidelines, that would be helpful, rather than the number of views and so on. The number of views is meaningless these days. Ms Lord : That's fine, of course. I'll just share some general information with you on notice about the YouTube Partner Program, but I would say that monetisation is complex beyond that. It really does become a question of how many times content is viewed. Then that's relevant to if that content is being actively recommended or being surfaced high up in search rankings on YouTube. As we've discussed, our priority there is around quality content and elevating that. The other point I'd note is that this question that we have some sort of financial incentive to surface poor quality content is somewhat misplaced. That doesn't work on our platform; we've learned that in the past. The reason that doesn't work for us is because it's just not in our business interest. That is because advertisers don't want their ads placed next to bad content. When we lose advertisers, we lose revenue. That also impacts the YouTube creator community, who don't have advertisers running ads against their content, and we have this revenue-sharing arrangement in place with them. Because of the way our platform operates, that's not necessarily the case for us. We do have a financial incentive to be a responsible platform. Senator ANANDA-RAJAH: Thank you. CHAIR: I've got a question that I'll follow up on a bit later, if I get a chance, but we don't have a lot of time. Senator Canavan, you've got 10 minutes if you'd like it. Senator CANAVAN: I'm not a regular user of your services, but I noted that particularly during the COVID pandemic, sometimes, when I put things up on YouTube or the like, I'd get a warning, because I'd speak about climate change. You had a thing about, 'Oh, you want to learn more about climate change?' or you put a label on me when I was against vaccine mandates and the like. Do you still do that? Do you still put these notes up, if someone speaks about a certain topic, that they'd need to go and find more information? Ms Lord : Yes, on YouTube we do, for certain topics. We apply these information panels on a range of different topics. That does include climate change. We consistently review and reassess which topics those information panels are surfaced on. Senator CANAVAN: I just checked. I don't seem to be getting the information panels anymore. Do you exempt politicians or something now or— Ms Lord : No, we don't. They're applied to the top of search results and on videos regardless of the speaker or perspective. It's just relevant to the subject matter. Senator CANAVAN: So you're saying I need to up my game. I took them as a badge of honour. When I see them on someone's post, I usually think, 'Well, that's something I want to read or listen to because it's obviously going against the establishment's delusions.' I'd love to see more of those, because I think they really help us identify good content that is actually challenging the groupthink that we live under at the moment. Can you tell me what topics your information panels relate to, at the moment—in Australia? I'm only really concerned about the Australian context. Ms Lord : Of course. It's topics that might be prone to misinformation or require some additional clarification. As I've mentioned, we have one on climate change. I believe we've got one on the Malaysian Airlines flight that went missing somewhere—no-one knows where—and on other sorts of things like that. Senator CANAVAN: Do you still do a COVID one or is that finished? Ms Lord : I don't believe so, but I couldn't confirm that. JJ, you're shaking your head. Mr Sahel : No, with COVID, we've moved on from that, just as others have. But we've got things like the moon landing, for instance, which remains one of those debates or sensitive questions. Senator CANAVAN: That would really be a badge of honour, if I could get one on the moon landing! Who does the information notes? Do you produce them, yourselves, internally? Ms Lord : That's correct. It's a product feature. It's a text box. It will contain a factual statement on the relevant issue and then a link through to either an authoritative organisation or something like Wikipedia or Encyclopaedia Britannica for our health labels. It might link through to the Mayo Clinic, for instance. It really depends on the topic. Senator CANAVAN: Do you put these information labels up on all content that might discuss climate change or is there some kind of stratification, triage of 'That content's a bit more concerning; we'll put the information label on that'? Ms Lord : No, we surface those information panels regardless of the perspective or the speaker. They'll appear at the top of your search results if there's a search related to climate change or about climate change, and then they'll surface in the video description as well. Senator CANAVAN: A number of people were banned from YouTube at the height of all this, particularly around the COVID issue. Are people still being censured for certain views that they might express on YouTube? Ms Lord : I appreciate that question. YouTube's a video library of over 20 billion videos. We're not here to police opinions, but we seek to provide a platform that's safe and responsible for our users to come to find video content that provides them with information to learn, to be entertained and for all sorts of reasons. We do have ground rules in place. They are our community guidelines, our policies. We are transparent about those. Everyone who seeks to upload content to YouTube agrees to abide by those community guidelines. They're publicly available for all our users as well. We hold ourselves accountable for how we're enforcing those community guidelines through a number of mechanisms, including quarterly reporting under our community guidelines transparency report, which we also make publicly available. And, of course, we do continue to review those policies to make sure that we are drawing the line in the right place and striking the right balance between protecting our users from harm and allowing space for discussion and the free flow of information. Senator CANAVAN: The chair is going to wrap me up— CHAIR: No, you've got more time. I just wanted to ask if Alex Jones is an example of someone that you'd banned previously— Senator CANAVAN: Yes, I think he had been. CHAIR: Is he still banned from YouTube? Senator CANAVAN: He's been proven quite right on a number of topics. CHAIR: I don't know about that. Senator CANAVAN: Well, he definitely has! Ms Lord : We can— Senator CANAVAN: Is Alex Jones still banned? Ms Lord : I don't have that information, but I'm happy to take that on notice, and we can confirm. Senator CANAVAN: He was guilty of exposing lots of misconduct; that's for sure. I don't know if I got an answer to my question before. I suppose I'll just reword it. Can people be banned from YouTube, in particular? Of your products, that's the one I'm concerned about. Can people be banned for expressing an opinion? Ms Lord : People can have their channels terminated in a few different circumstances. The first is when they have had three strikes against the same policy in 90 days. The second is where they're running a channel that's wholly dedicated to violations of our community guidelines. The third time that we will terminate a channel is in a single severe instance of abuse. We're clear about that. That's publicly available information. Senator CANAVAN: I'm not sure if you've quite answered me. I get the abuse—definitely happy with that—but what about just expressing an opinion? If I express an opinion on a particular political topic that falls foul of some standard, can that contribute to one of my three strikes? Ms Lord : As I mentioned, the strikes are based on violations of our community guidelines, and they're clear and transparent for all our users to see. Senator CANAVAN: I suppose that's what I want to know: do those community guidelines go to the expression of opinion? Ms Lord : I can share the community guidelines with you. They're publicly available, and, if you're uploading content, you would have agreed to follow those community guidelines. They cover— Senator CANAVAN: I don't read all that legalese. Most people don't. Ms Lord : Let me share our community guidelines with you— Senator CANAVAN: I don't understand why you can't answer that question. It's a very simple question. You obviously, presumably, know your own community guidelines. I don't know why it's my responsibility to be familiar with yours. Can you be banned? Can you be banned for expressing an opinion? Ms Lord : As I said, we're clear and transparent with our community. That includes people who are uploading content on what we don't allow on our platform. Those community guidelines are publicly available, and anyone who's uploading content to YouTube agrees to abide by them. We're also very clear, with our creators, around the process for enforcement of those guidelines. I think we strike a good balance there. You would get a warning. Initially, you would have certain privileges suspended for a week, and it's a scaled enforcement up until the point that, if you've got three strikes for the same policy violation, then, yes, your channel would be terminated. Senator CANAVAN: I don't think I'm getting an answer, Chair, so I'm happy to pass the call over. CHAIR: Senator Pocock, you have the call. Senator DAVID POCOCK: Thank you for your time this afternoon. Do you accept that AI tools, like Google's Gemini, present a significant challenge in mis- and disinformation being created and spread? Mr Huntley : I can take part of this. I certainly see that AI is an incredibly powerful tool, and we are kind of on the enforcement side. I think it is a tool that can be greatly used to fight against these threats, but, as a powerful tool, it can also be used by the bad guys as well. We've been very transparent about what we're seeing currently from serious threat actors and how we're seeing this abuse. We're also investing very heavily in detecting this abuse, preventing this abuse in our systems and detecting it. Every time we've had a technological leap, we've actually seen threats evolve. They evolve to warfare, to crime. We are certainly of the position that we are going all out to use this capability, and we used it all the way back from—we've used AI for spam filtering for over a decade now. We want to be on that forefront. We think that AI can actually allow us to win against these threats and make things better, but it requires a heavy investment for us to use that. But, certainly, this tool, like all tools, will be used by the bad guys. Senator DAVID POCOCK: What I heard is that you accept that it presents a serious risk but it also presents an opportunity and you're working hard to minimise the risk. Mr Huntley : Yes. Senator DAVID POCOCK: You said that you're transparent around threats. What evidence is there that Gemini has been used to create content that has mis- or disinformation or has been used to direct bots? Mr Huntley : For content produced on Gemini we have a technology called SynthID, which is built on, which puts in watermarks and systems for detection. We think that Gemini is one of the harder platforms to use for this particular type of abuse, and we have strong detections across our platforms. This is an industry standard that we've developed. We certainly have seen limited use of these capabilities to develop phishing messages, to research how to target people. We have seen some levels of abuse and we have produced two reports from our threat intelligence group. We can provide those afterwards about the specific details of what we've observed of abuse on our platform. We certainly have people dedicated now, as one of our highest priorities, to detect this abuse and surface these threats, because it is an area of great concern, but also we need to work together to fight these threats, so we're trying to surface that as much as possible. Senator DAVID POCOCK: How many people are dedicated to looking at this issue? How big's the team? Mr Huntley : I suppose it's where you define the limits of this issue. I would say that the AI threat generally is a huge theme across multiple teams, including the intelligence group, including trust and safety in YouTube, in DeepMind— Senator DAVID POCOCK: I understand that. So maybe— Mr Huntley : There would have to be 100 people, at least, touching this. Senator DAVID POCOCK: I understand that there'd be hundreds across a big company like yours. But let's say we're specifically looking at mis- and disinformation and bots. How many people would touch on that? Mr Huntley : I'd have to take that question on notice to get a solid number. It would be a fairly large number spread across a number of teams. Senator DAVID POCOCK: Thank you. Given what you've said about the very serious risks but also the opportunities, can I take it that you would be supportive of safeguards or guardrails to protect against the use of AI to create mis- and disinformation? Mr Huntley : This is a highly evolving space and we have to be very careful, when it comes to policy legislation, that we are adaptive and that we don't lock ourselves in. I do think we are very quickly adapting, working against these threats, so it's about making sure that, whatever we do in this space, we don't lose that dynamism and ability to do this. Senator DAVID POCOCK: How would safeguards or guardrails against using AI to create mis- and disinformation lock us in? Are you supportive of safeguards or not? Mr Huntley : I'll hand over to my policy colleagues here. I'm more on the technical side. Mr Sahel : I may just offer something. Thank you for these important questions and these are the questions we've been grappling with for a while now. We came up with AI principles back in 2017, when we were already thinking about what sorts of guardrails we should put in place, and we're working with policymakers here in Australia and beyond to try and think about what would be the right approach. One way to think about it, when we think about technology—Shane was mentioning how we've developed something called SynthID, which basically tags each image that's generated by AI so that it can be then analysed. The tag can never go. It's a mark that can't be tampered with, we believe. What we would hope is that those sorts of marks that help to identify AI generated content are not just generated by Google but are adopted across the industry so that we have an industry-wide set of preventative tools that can then be translated into helping users to actually understand if content has been generated by AI and, therefore, whether they need to pay more attention to this. I think it's about finding the right sort of technical means and then backing that up with wider practice. We're hard at work on exactly that at the moment, both within industry and in partnership with policymakers and around the world. Senator DAVID POCOCK: My concern—and I think this is shared by many Australians—is that we are being warned about the not insignificant risks that AI poses to us on a whole range of fronts. We hear Google, Anthropic and others talking about some of the very concerning developments that are coming up as you are monitoring how these AI systems are developing and organising. Yet, every time we talk about guardrails, we hear about the tools. 'Don't worry. We're looking into this ourselves and we're developing these tools.' I think as a society we should be saying, 'Sure, you can work on your tools to make your AI safer, but here are the guardrails that we as a society are putting on you.' I'm interested in whether or not you agree that there should be guardrails on AI development. Or, do you think, like OpenAI may believe, that it should just be, 'Do whatever you want'? Mr Sahel : We are definitely in favour of guardrails. I think, a bit like what Shane was saying, we just need to make sure that we devise systems that are flexible enough to evolve as the technology evolves. As I said, we're involved with policymakers. We've supported bills in the US, for instance, around things like what we call 'digital replicas' to try and put in guardrails in terms of how AI generated content is produced and displayed. So we're for it. It's not a question of whether we need guardrails; it's what the guardrails should be et cetera. We're very happy to continue to engage with you and other lawmakers in Australia on that and happy to share more details on our exact positions and suggestions on notice, afterwards, in writing, if you would be interested. Mr Huntley : Certainly, Kent Walker, our senior vice president, has said: AI is too important not to regulate, and too important not to regulate well. In 2024, we published seven principles about how to get AI regulation right. I think that is incredibly important. It's also incredibly important in an Australian context. We want to make sure that we do counter the threats of AI, but we do need to make sure that Australia is able to harness the great opportunities of AI as well. That is why we are for regulation. But we have to get the regulation right. Senator DAVID POCOCK: My concern is that Australians hear that, and it's like: 'Yes, we're for regulation but not now. Let's take time to get it right.' At the same time, all of the big AI companies are progressing at a rate we've just never seen in human history. The kind of progress we're seeing—we're seeing job losses across the board; a whole range of concerning things coming up. I think this is something that the parliament is, ultimately, going to have to put on companies rather than taking the next five years to slowly work out. I really do appreciate your time here today. I wanted to finish, if possible, on one question that has been raised with me a lot. I know this is potentially at the very edge of our terms of reference, but the community I represent here in the ACT, and many Australians around the country, are concerned about the climate and environmental impact of data centres. We're told that in Sydney alone they'll be using more drinking water than the entire capital city of Australia in the next decade. What's Google doing to ensure that there isn't a spike in emissions, electricity consumption from the grid and water consumption? Ms Lord : JJ, are you right to take this one? Mr Sahel : Google's been committed to an environmental agenda for a number of years now, and that certainly extends to data centres. We're very aware of all this, and we're constantly attempting—across our value chain, if you will, from the servers all the way to data centres—to understand and minimise the environmental impact of what we do and move towards being a carbon-neutral company. We're well on the way to these goals. We would be happy to provide more specific numbers, but we're very aware and share the concerns for the need to make sure that, as our tech consumption continues to increase, we minimise the environmental impact. In fact, the trend is positive, because we're finding ways in which we can minimise the negatives. It's trending in a good direction, and we're happy to take on, just to be careful— Senator DAVID POCOCK: That would be great. Senator ROBERTS: Thank you for appearing today. I have several questions. I'll give you some context to the first and most important: to me, human progress has been enabled dramatically in the last 170-180 years by two things—freedom of thought and freedom of sharing one's ideas and building on someone else's ideas. Secondly, responsibility comes from choice. Without choice then it's a dictatorship. People are turning to YouTube for better education; is that correct? That's my understanding, because they're sick and tired of woke universities and they know they can get practical experience by just educating themselves on YouTube. That applies to high school students and uni students. Would you agree? Ms Lord : I can't comment on the motivation, but yes, we've had a number of data points that have confirmed that teachers, parents and children alike find value and educational content on YouTube. Senator ROBERTS: So who are your competitors? University lecturers, high school teachers? You must have a plethora Ms Lord : That's a really interesting question. I haven't thought about those sectors as our competitors, to be honest. I would imagine that, rather than us being competitors, our videos are useful for them and our platform is a tool for them to share their knowledge with users. I can think of a number of teachers like Eddie Woo who create educational content, or content creators like Tibees, who was doing a PhD in astrophysics and now runs a channel with maths, physics and astronomy content, and is wonderful. I don't know that we'd see ourselves as competitors, but more as a platform that provides them with an opportunity to reach audiences and that provides users with the opportunity to access high-quality information. Senator ROBERTS: What keeps your audience engaged on YouTube? Ms Lord : That's a really good question. As we've discussed, we are a video library. We have a corpus of over 20 billion videos on our platform. That comes from a wide range of sources. When we recommend content to users we seek to prioritise quality for them, and those recommendations are based on a number of signals that users shared with us that are designed to provide information to them that's actually relevant and useful. Let me give you an example: if you were to come to YouTube and search for 'bulldogs', based on signals that you would have shared with us in the past, that could mean that we would be serving you content on a football team or content on dogs. There are a number of reasons people come to our platform. Senator ROBERTS: You said, and I think these words are a quote, 'best content is in your own interest'; correct? Ms Lord : Given the vastness of our content library, the recommendations and quality rankings assist users to find content that's useful and relevant to them. Senator ROBERTS: If you provide lousy or threatening content then advertisers are not going to be associated with those content streams, are they? Ms Lord : As we answered previously, we have a business incentive to ensure that our content is of a high quality. Advertisers are not interested in having their brands associated with harmful content. Senator ROBERTS: Except for child abuse and child exploitation, except for terrorists, except for advocating violence and harm, surely, you're the best adjudicator. Those things, I think, should be off. Surely you're the best assessor, because it's in your own interest to provide the best content. Correct? Ms Lord : I'm confirming that the types of content you've just referred to are not allowed on our platform under our community guidelines. Senator ROBERTS: Thank you. I wasn't inferring that. I was confirming that they're not— Ms Lord : I understand. I just wanted to be clear that that's not permitted. Senator ROBERTS: So advertisers, generally, would not be looking straight for clicks based upon harmful content? Ms Lord : I'm not sure I understand your question. Would you mind just— Senator ROBERTS: What's the incentive for you? Is it to put out good content or is it to put out clickbait that can get clicks and advertisers? Ms Lord : Our incentive is quality content, and that is because it attracts advertisers but it also retains users. It's what people come to look for, high-quality content that's useful and relevant to them. Senator ROBERTS: The definition of quality—I did some work on this 20 years ago—is a 'predictable degree of uniformity and dependability at low overall cost to the customer'. Let's move on from that. CHAIR: This is your last question, Senator Roberts. Senator ROBERTS: You mentioned, in regard to climate change, that you're linked to a page to the United Nations for checking. What page is that, please, and is it a page of the United Nations Intergovernmental Panel on Climate Change? Ms Lord : I'm happy to share that link with you so you can take a look for yourself. It's just a general information page, on climate change, hosted by the United Nations. Senator ROBERTS: Could you, specifically, show me or identify where the science is on that, the empirical scientific proof? Ms Lord : Let me share the page with you and you can take a look at it yourself. I'm happy to share that with you. Senator ROBERTS: Thank you. CHAIR: Thank, Senator Roberts. If you've got any other questions, please put them on notice. I have a couple of quick ones. I raised the same report with both the other two digital platforms we had, Meta and TikTok. It's from the Center for Countering Digital Hate. The report's called, Extreme weather: how a storm of false and misleading claims about extreme weather events spread unchecked on social media putting lives at risk. I'm not sure if you had a chance, Ms Lord—and you may have seen the previous hearing—to have a look at it. If not, I'll ask you to take some questions on notice, on this report, and I can speak about it generically. In the report, the case studies are from 2024 and the beginning of 2025. Essentially, they look at misleading claims about the causes of extreme weather events, misleading claims about disaster relief aid, misleading claims about emergency responses, misleading claims about the impact of climate change, extreme weather events, and misleading claims about political responses to extreme weather events. And they do look at a number of platforms. There are two things I want to ask you about that they pulled YouTube up on. First is that 29 per cent of identifiable false or misleading claims, misinformation videos, around these extreme weather events did display ads. Of course, it's against your policy to have advertisers with verifiably misleading content. Do you have a special team that will look, for example, at fact-checking content that goes with ads? I know you don't do it across the board; you have information panels or community notes. How do you do that? What kind of assurance can you give us that we aren't seeing advertising alongside climate misinformation which has been outlined in this report? Ms Lord : Sorry, I'm not aware of that report and I haven't had a chance to look at it yet. My apologies. I must have missed it in the previous hearings. I'm happy to take a look at that and then take any questions on notice. CHAIR: Yes. There are a number of claims in there, and I can put some more detailed questions. In terms of operationalising this, do you have a different set of rules with advertising of higher standards—how do you fact-check posts where there is advertising attached to it? Ms Lord : Our organic content on YouTube is obviously subject to our community guidelines, and I can share those with the committee. You can take a look at those. JJ, I don't know if you would be able to speak about some of the ads policies as well that we have in place? Mr Sahel : Perhaps just to a high level and we can come back in more detail on notice, as Rachel suggested. CHAIR: Thank you. Mr Sahel : We have a range of policies that are specific to our advertisers or publishers of content on our platforms, across all Google platforms, not just YouTube. These include policies of good behaviour, such as policies against unreliable claims and impersonation, for instance trying to make sure that an advertiser is not pretending to be a qualified professional in a particular sector when they're not and trying to advertise or sell a particular product on that premise or, indeed, that they are not making unreliable claims in order to promote a particular point or product that is based on information that is demonstrably false according to international consensus. CHAIR: Let's leave it for on notice, but, if you look at page 17 of the report, there's an example there of a Charlie Kirk post around an extreme weather event that has Starbucks ads on it. It clearly looks like misinformation to me, but I'll let you come back to that. Second, another thing they pulled YouTube up on is that nearly one in three YouTube videos that they studied that were promoting misleading claims about extreme weather featured a recommendation for further climate sceptic or climate denial content in the Up Next panel. They gave the example that, of the 100 videos they looked at, 30 had at least one recommendation contained within the top five recommendations which contained false or misleading evidence of extreme weather events or climate denial. I suppose this question comes down to your algorithm. By the way, I love a lot of your content. I was looking at something about the Roman general Sulla last night and how he configured some of his battles. I do like being shown other things in that area. I understand it's a two-edged sword here. Do you accept that the algorithm does push people down rabbit holes and into echo chambers if they continue to be shown similar content to what they've clicked on? Isn't this a really big problem for us getting the right information to the people who need it the most, in terms of countering climate disinformation? Ms Lord : I really appreciate that question. I also appreciate the positive feedback about YouTube, so thank you for that. As we discussed, the purpose of our recommendation system is to find videos from our very large corpus of 20-plus billion videos that are relevant and useful for users. We draw on a number of signals to do that. We give users control over recommendations, including the ability to turn them off on the home page and to shape recommendations as well. There are a number of controls and tools for that. What you're getting at there in your question, I think, probably misunderstands how we use recommendations on our system, on our platform. From our perspective, recommendations are actually a vital part of how we maintain a safe and responsible platform. We use our recommendation system as a tool to prevent potential harms. As we've discussed today, we are really focused on elevating and prioritising quality content. Topics where accuracy is really important—and that includes science information—will surface content from a range of sources but also feature a range of perspectives. I think that's a recommendation system that's working well. So, yes, you might see some content that carries one perspective. You might see content that carries another. It might be a discussion of the issue. It might also be a parody or a comedy sketch on that topic. But what we're looking at there is diversifying the recommendations that people are seeing. Of course, if content violates our policies it's removed and it can't be recommended at all. CHAIR: There's other stuff in there—that 73 per cent of the posts that did have misleading information around extreme climate events weren't from verified accounts et cetera—but I'll ask you, on notice, if you could provide a response to this report. The committee will get you a copy of the report. I'll go to Senator Ananda-Rajah for a couple of quick final questions. Senator ANANDA-RAJAH: You did allude to this earlier with regards to your watermarking of AI content, but, in general, how is AI being used in your platform to detect and prevent the spread of false or inauthentic material? How accurate do you think your tools are in achieving that? Ms Lord : I'm happy to take that question for YouTube and share some information with you about how we're using automated detection generally to pick up policy violative content. As I've mentioned on a number of occasions today, we report quarterly on our enforcement of our community guidelines. That's publicly available, and we will share that with the committee. As part of that report we talk about first source of detection. Let me break down some figures for you. Of over 8.6 million videos we removed in the last quarter of last year, 97 per cent were first flagged by machines, demonstrating the value of those tools. That meant 42.3 per cent never received a single view, and a further 43.3 per cent were only removed with one to 10 views. From our perspective, what we're seeking to measure is—I know we focus a lot on turnaround times. We focus on how widely the content's seen, because we could have a video sit on YouTube for 10 years and be seen twice or it could be there for five minutes and be seen 100 times. That's how we hold ourselves accountable. The other data point that I think is worth sharing with you is our violative view rate. This is also included in our transparency report and updated quarterly. That is a measure of the number of views that come from violative content on YouTube. For us, it's a measure of how well our systems are working, and that obviously includes automation. In that same period, of the fourth quarter of last year, for every 10,000 views on YouTube only 14 or 15 came from content that violated our community guidelines. Senator ANANDA-RAJAH: In the ACMA report there's also mention of your academic research awards that you awarded, in October last year, to two projects on misinformation at Monash and at UTS. If you could supply us with those projects on notice, when they're finished, that would be great. Can you also talk us through AI search now? It's clearly dominating traditional search. What is being done to ensure authoritative information is preferenced in Gemini, for example? What is going to happen with this issue around advertising, if people just preference the AI search and stop looking through the rest of the content in Google? Mr Sahel : Perhaps I can answer this many layered question from a couple of angles. I think the first one is around information quality, which is at the heart of Google's mission and the way that we've developed our products. When it comes to search, as we've explained, I think, throughout this hearing, we try to surface the highest quality information based on a variety of characteristics, such as the reliability of the source, the reputation of that source and other such factors. That's the sort of approach which has guided search for 20 years, and results an AI overview would provide would be guided by that as well. That's the same philosophy, if you will. As for development, there's a lot of information that's surfacing where we're seeing some positive results for publishers, in terms of the evolution of AI mode and overviews in terms of their revenues. But I think what's really important to keep in mind is that we're still in very early days. As I think Senator Pocock was mentioning earlier, this is moving very fast, and both the technology and the business models are evolving very fast. What we tried to do very early on, as we created Gemini and search results aided by AI, was have attributions and links to the particular content rather than just provide an answer, so people can look at the detail. We provide links directly next to the answer to some of the key sources. It's very much part of this approach of information quality, giving the user broader context and a variety of sources. We're continuing with that sort of overall philosophy, but I think the technology is moving very fast and we'll have to see how it moves. This is very much top of mind for us. Senator ANANDA-RAJAH: Could you provide us with a list of the criteria you used for information quality and perhaps a sense of how you weighted each of those criteria. That would be useful. Mr Sahel : Yes. We publish transparently what we call our search quality rater guidelines. We're very happy to share that and related information on notice, of course. Senator ANANDA-RAJAH: Does that also apply to YouTube? Do the same criteria apply to YouTube, or is it a bit different? Ms Lord : We identify multiple signals that help us determine channels that demonstrate expertise and trustworthiness on a given topic. But, yes, we also use the search quality rater guidelines—sorry, I find that quite a mouthful—to help us assess the quality of videos on YouTube. We can build that into the response that we'll provide on notice. Senator ANANDA-RAJAH: Thank you very much. CHAIR: I have a very quick question for you, Mr Huntley, if you don't mind. In relation to security, does Google work with intelligence agencies in places like Australia in relation to coordinated inauthentic behaviour and foreign interference and foreign state actors? Mr Huntley : Yes, we engage with governments worldwide, including the Australian government, on these topics. I've taken meetings and briefed on this topic. We engage with them regularly—with governments around the world. CHAIR: Does Google, across the board, have granular data on take-downs, for example, around potential foreign interference or even other forms of online harms? I know there are some reporting requirements under the code here in Australia. Mr Huntley : Certainly. On that policy, for quite a number of years now, we've produced what's called the TAGBulletin, which outlines, specifically, all the campaigns that we have seen around coordinated inauthentic behaviour with the counts of videos and some details around them. This is updated in detail and is comprehensive about the threats we've taken down under that policy. CHAIR: Are there any restrictions on researchers? Obviously we've had a lot of researchers provide feedback and submissions to this inquiry—and evidence. Are there any restrictions on researchers accessing this kind of information? Mr Huntley : It certainly depends on what the type of information is. For certain people, like researchers, we do provide some information, but certain information is obviously either proprietary or privacy related information that we can't provide to outsiders. Someone else may want to talk about what we do provide. CHAIR: For example, under the EU's Digital Services Act, you've got requirements to provide access for researchers in relation to some— Ms Lord : I'm happy to jump in here. Sorry to interrupt you. CHAIR: We don't want to be taking you over time, by the way. We really appreciate your cooperation with the committee. It's greatly appreciated. Ms Lord : It's not a problem at all. We do have a YouTube researcher program, and that provides scaled access to global video metadata across the entire YouTube corpus. That is open to academic researchers who are affiliated with accredited higher learning institutions. I'm happy to share through some further detail with the committee on that. As you've seen in our transparency report under the Australian Code of Practice on Disinformation and Misinformation, we have researcher awards in Australia. JJ is probably better to speak about some of these things than I am. We make publicly available data around trends on Google and YouTube. JJ, I'll pass to you if there's anything else I'm missing there. There's a lot of transparency for the public and then specific access for researchers and support for the research community as well. CHAIR: Great. If you could provide that to us, that would be fantastic. Thank you so much for your time here today. JJ, I do like the octopus behind your head—the picture in the frame. In relation to information provided, because the committee's inquiry is going to be wrapping up very shortly, the committee has agreed that responses to any questions on notice should be provided within one week of witnesses receiving a copy of the Hansard transcript. That's usually a couple of days, but we'll clarify that with you. That concludes today's hearing. Thank you to all witnesses who appeared, to Hansard and Broadcasting for their assistance and, of course, to the wonderful secretariat who worked so hard on this inquiry. Committee adjourned at 17:44