Apologies have been received from Deputy Ó Snodaigh and Senator Noonan.

Today's meeting has been convened to continue the committee's consideration of the regulation of online platforms and supports to improve online safety and participation. I welcome the following witnesses to the meeting: Ms Susan Moss, head of policy and government relations at TikTok; Mr. Richard Collard, minor safety and public policy lead at TikTok; Mr. Ryan Meade, government affairs public policy manager at Google; Ms Chloe Setter, child safety public policy manager at Google; Mr. Dualta Ó Broin, head of public policy at Meta; and Mr. David Miles, safety policy director for Europe, Middle East and Africa at Meta.

Representatives from X were also invited to attend today's session but declined. This is incredibly worrying. It is extremely disappointing that a company whose European headquarters are literally down the road from here has refused to attend an Oireachtas meeting to discuss issues with which it has been closely involved. Not alone did the committee write to and communicate with X on a number of occasions, but I asked the Taoiseach to write to the company. He has confirmed that he did so and yet X, as a company, still refused to come before the people of Ireland, the Oireachtas and this committee, which has jurisdiction over this area. I have asked the Taoiseach for the correspondence with X but it has not been provided, which I find a little bizarre, but I am sure it will be provided at some stage. It is deeply worrying, dissatisfying and pretty disrespectful that an organisation, an e-commerce company, of the scale of X is not here today to discuss these issues.

I thank the representatives who are here for coming. There was a little bit of conditionality in the replies we received from the companies, basically saying that if three of the four companies turned up, they would turn up. I found that extremely strange. Actually, I found it unacceptable, to be honest. Witnesses cannot impose conditionality like that on an Oireachtas committee. They are here as representatives of their own entities. In another sphere, people would wonder why companies would want other companies to be here or not to be here. It is extremely strange and worrying.

Before I move on to the opening statements, I want to outline the conditions under which this committee works. Witnesses should note that I will be using the terms "e-commerce companies" and "publishers of content" rather than "platforms" because that is what the companies are. The committee believes that the witnesses are here to represent their companies corporately. That means corporately, full stop. Their answers to questions here will be taken as representative, corporately, of their organisations. Furthermore, witnesses will be aware that this meeting is taking place within the jurisdiction of Ireland and therefore the principle of adherence to Irish law is very important.

Generally, the format of our meetings is that I invite witnesses to deliver an opening statement that is limited to three minutes. However, given the interest in today's meeting, we will be taking the opening statements as read because we need to have all of the time available for a discussion. Is it agreed that the statements will be published on the committee's website? Agreed.

Before I open the floor to questions, I will clarify some limitations in relation to parliamentary privilege and the practices of the Houses as regards references witness make to other persons in their evidence. The evidence of witnesses physically present or those who give evidence from within the parliamentary precincts is protected, pursuant to the Constitution and statute, by absolute privilege. This means that they have an absolute defence against any defamation action for anything they say at this meeting. However, witnesses are expected not to abuse this privilege and it is my duty to ensure that this privilege is not abused. Therefore, if their statements are potentially defamatory in relation to any identifiable person or entity, they will be directed by me to discontinue their remarks and it is imperative that they comply with any such direction. Members are reminded of the long-standing parliamentary practice to the effect that they should not comment on, criticise or make charges against a person outside the Houses or an official either by name or in such a way as to make him or her identifiable.

Finally, before we begin our discussion, I want to welcome several transition year students from various institutions to the Public Gallery. We are joined by Daniel, Noel, Leah, Arianna and Switzlov, all of whom are very welcome. Deputy Joanna Byrne will now begin proceedings.

I thank all of the witnesses for being here. It is a great start that they are actually present in the room. It is fair to say that the big tech platforms will always say that they are doing everything they can to ensure online safety and privacy for users.

However, our own Data Protection Commission, DPC, has issued over €4 billion in fines over the last five years to online companies, including €530 million last year. To date, it has only collected 0.6% of those fines. Does the fact that those fines have been issued in the first place not make a mockery of the opening statements? Could the three entities declare whether they have any outstanding fines today, and if they do, when they intend to pay them?

I ask all three witnesses to respond quickly.

We have an ongoing case in the High Court, so I am limited in what I can say in that regard.

Could Ms Moss reply to the second part of the question?

No, that is okay.

My understanding is that we do not have any fines levied against us at this time but the point the Deputy makes demonstrates that GDPR, which has been in place for a number of years now, has resulted in a significant amount of investigations across the board.

Yes, we do have a number of cases before the courts in relation to fines, judgments and findings of the DPC. It is tempting to look through the lens of fines as to whether a regulation is effective or not, but I do not necessarily think that is the best way to look at it. If we look at the changes that companies have made since the GDPR was introduced and the programmes that we have put in place, that is a far better metric to determine the success or otherwise of the regulation. Yes, we do have fines.

They were €530 million in the last year alone. That is a substantial amount of fines, which suggests the regulation is not as effective as it could or should be. We might move on.

I want to speak a bit about algorithms. It is something that has come up in every sitting in recent weeks when we are discussing this topic. Researchers have done tests on TikTok and YouTube. One case I am going to reference is a simulated account of a 13-year old who watched one video containing eating disorder content. After that, they analysed the results of 1,000 video recommendations by YouTube and found that one third were for harmful eating disorder content, including videos endorsing extreme calorie restriction and promoting dangerous diets like anorexia boot camp. These videos are in breach of YouTube's own policies but they were still there.

What we take from that is, when tested, the platforms are failing. They allow harmful content, suicidal ideation and harmful eating disorder content. With all that staring us in the face, 300 workers who worked in content moderation are being laid off as Meta moves to AI. Why is Meta replacing human moderation with AI to tackle this? It is a really important topic that is affecting so many young people in particular at this time. My presumption is that the reason that change is being made is, first, to deflect blame or, second, to cut costs. Who is going to be responsible if and when AI fails?

There is quite a lot in that. Just to unpack it, on the recommendation system and how the algorithms work, there are regulations in place that cover the use of algorithms and ensuring that unacceptable risks are not being created by the algorithms. That is provided for in the Digital Services Act, DSA, for example. We have a range of controls in place that users can use to control their algorithm, for example, changing it to just having friends and followers' content, which would just be a chronological feed of content rather than having a suggestion by the recommender system working for a user.

The most powerful thing we have done in that space is to create a separate experience for under-age users, namely, teen accounts. Users between the ages of 13 and 15 are subject to the most restricted levels of content and the most restricted privacy settings. Above that, users have a little bit more autonomy until they become an adult user. It is about putting in place those controls to ensure that users have an age-appropriate experience. That is critical to dealing with it.

The suggestion is that recommender systems are a kind of binary thing and that if we could turn them off, everything would be fine, but the reality is that it is far more complex and nuanced than that. Ensuring that the right degree of control is there for users and parents is a really important part to play.

I assume the Deputy is referring to the Covalen workers. The requirements that we have in relation to outsourced partners will change over time, depending on particular needs, locations and languages. We will always be looking to ensure that the right mix is in place between AI and human review. That is where we will be at, ensuring that we have the mix right to ensure that we get the most accurate results.

Is Mr. Ó Broin confident that Meta has the mix right at this moment in time?

Yes.

Do the other witnesses agree with that?

To speak to content moderation and AI, we have used AI machine learning in content moderation for many years. We have been continuing to invest in both the AI classifiers that we use but also in our human reviewers. They work together. We have seen big improvements in the AI systems in certain categories of content. For example, the vast majority of the violent extremist content that is removed for violation of our policies is first flagged by an automatic system. We always have a human in the loop, and we have escalation to human reviewers as well. I do not think it is a case that we are in any sense replacing human reviewers with AI; they work together and they have done for many years. It has been to the benefit of our users because, as the AI systems get trained on real-world examples, they get more accurate at flagging content to be removed.

They are still open to manipulation.

In what sense?

In how the algorithms are targeting specific topics that might be relevant today. Algorithms can-----

Just to give an example of how we work with humans and machine learning, automated classifiers look for content that it has the confidence level to indicate may be violative of our policies. Once that is human reviewed, the accuracy of the model can be fine tuned based on the human's reading of the policy. It is a virtuous circle in that sense, in that the combination of the AI and the human allow both to work more effectively.

I thank Deputy Byrne very much and I call Deputy Carrigy.

The witnesses are very welcome. I will choose a line from each of the statements to put to them: "there's no responsibility more important than protecting the people on our platform"; "We have robust policies against illegal and harmful content," and "in particular young people, don't have to see content meant to intimidate, exclude or silence them." That is what the companies say they are going to do but the reality is that is not happening. Otherwise, we would not be sitting here today discussing this issue, in particular the access our children have to these platforms and the illegal content that is on them. Who do the companies believe is responsible for online content? Is it the companies' responsibility or that of the Government? Could a witness from each company please respond?

The user that posts the content has the responsibility in the first instance for putting content on the platform, but it is the responsibility of the companies to ensure that users have an age-appropriate experience. For example, in the case of teen accounts, it is that young users do not see content that is harmful to them.

Could we have the next response, please?

I absolutely agree that we have a responsibility as platforms. If we want to have customers and users that include young people, then we have to make sure that we put in place age-appropriate experiences and build for young people, which is something Google has been doing for decades. We need to enforce against our own policies, which is what we do through content moderation, and also be transparent about that through our transparency reports and our DSA risk assessments in terms of what those risks are, how we are tackling them and how we are mitigating those risks.

It is the responsibility of the platforms to ensure that systemic risks are not apparent on our companies and our platforms and the content that is being seen by individuals.

That is our legal obligation under Articles 34 and 35 of the Digital Services Act. As companies, we are all independently audited for risk assessments under the EU Act and those reports are published annually for committee members to read. I recommend reading some of those reports because they will see the transparency and the work we do on the very dangers the Deputy is referring to.

Does Ms Moss accept the companies are not getting it right, yes or no?

No, I do not accept that. The presence of problematic content does not necessarily mean companies are failing. Our responsibility is to ensure we are minimising harm, that is, bringing it as close to zero as possible. Our responsibility really lies in getting that content down as quickly as possible, but its mere existence does not necessarily mean failure. It means we have to do better with enforcement.

The companies need to do better.

Deputy Joanna Byrne mentioned fines that have been imposed since Coimisiún na Meán was set up. Members dealt with it in the Online Safety and Media Regulation Act in the last Dáil. We looked for an online safety commissioner to have a role in that organisation. Why are the companies not paying the fines that have been imposed on the company for not adhering to the rules in place?

Google has not been levied with fines under those regimes to date, so I cannot speak about that.

There have been no investigations under the Online Safety and Media Regulations Act on Meta.

Have fines been issued?

No fines-----

Have any been issued on any of the three companies?

I think TikTok mentioned that there are moderators to moderate the content - technology moves - and using AI technology to remove personnel who can view content to see if it is harmful. Does Ms Moss have full confidence in the use of AI technology to review this, rather than people, who I feel should do it?

It is not a binary choice. It is not one or the other. Humans remain an extraordinarily important part of content moderation at TikTok. More than 86% of content is proactively removed by our AI models. We use them to support our human moderators and people who work at the company. Staffing levels may change over time based on operational needs and systems improvement but what is not changing is our investment in safety and people. We use AI for scale and humans for much more complex and contextualised issues such as misinformation and political advertising. However, humans - or people - will always play a critical role in moderation at TikTok.

What can each company say it will do in the future to eradicate this content? What are their plans, using technology and the significant finances available to them, to eradicate what is harmful to our children? At the end of the day, that is what this is all about. There are some good things, but we are facing a generation that is disseminating false information and significantly harmful content. Organisations and people who perhaps use other platforms are also advertising on the witnesses' platforms to try to direct traffic. I saw something recently about pornography where OnlyFans people are advertising on some of the witnesses' platforms and the whole idea of it is to drive people to pornographic material. That was in an article by Eoghan Cleary.

On the Deputy's question about how we intend to tackle this, we look at three main methods. One is continuing to develop and enforce our policies so we have the right policies against content that should not be on the platform. Then we have the investment-----

Even though they are not working properly at the moment.

Our transparency reports show the extent to which they work. For example, on YouTube, the vast majority of violative content - that is content that violates our policies - is removed before it reaches a handful of views. That is a measure we track on a quarterly basis and our internal teams are incentivised against that measure. We take seriously keeping that volume very low. However, we also recognise the need to create age appropriate experiences. For example, we developed YouTube Kids. For older children, we have supervised experiences, which allows parents to decide the extent to which young people have freedom to choose the types of content they see.

The third thing we do is invest in partnerships. In Ireland, we have long-standing partnerships with Barnardos and other organisations to promote the Be Internet Legends curriculum, which is a foundational online safety educational curriculum for young people.

It is a multipronged strategy. It will never be fully complete. We have to, on an ongoing basis, continue to improve and report on those improvements to be transparent about how we are doing.

I thank the witnesses for coming. I share the Cathaoirleach's annoyance about X's non-appearance before this committee. It is less Hamlet without the prince and more Othello without Iago.

I ask each of the companies to give a simple "Yes" or "No" answer as to whether they think X should be before the committee today.

It is important that companies take the opportunity to appear before the committee. I also think that, with the issues we are discussing, there is perhaps a broader range of companies.

I appreciate that, but it is a "Yes" on X.

Yes, as a business, TikTok believes in transparency and accountability and that is why we are showing due respect to the committee today.

I take that as a "Yes". I accept there is a broader range of companies.

I appreciate the work the witnesses do as individuals and the fact their companies have signed up to the online safety code, but the challenge is that we are looking at issues around online safety here again. We heard testimony, as the witnesses will be aware, from Gráinne Seoige recently. At the Committee on Artificial Intelligence, we have been dealing with some of the issues regarding Grok. The witnesses will be aware of some of the discussions we had around the Online Safety and Media Regulation Act and a lot of the debate on online safety. The reason we are here is that we know from listening to our constituents, including parents, children and young people, that they are concerned that the measures in place at the moment are not sufficiently strong.

When a company roles out any product or service, it tends to have two approaches. One is safety by design, which means right at the start it makes sure all of the product, before it is rolled out to consumers, has all the safety measures in place. Then if it is rolled out and a problem arises subsequently, the company calls back the product or puts the measures in place to rectify it. The fear we have across the sector is that companies are not responding sufficiently. Can the witnesses assure us about how that is done?

My colleagues spoke about fines. The witnesses will be aware from during the debate on the Online Safety and Media Regulation Act, that I take a strong view on personal responsibility. I welcome the statement by the Spanish Prime Minister, Pedro Sánchez, yesterday that we should move to a situation that where a company knowingly and willfully allows a dangerous or harmful product or service to be rolled out on its platform, individuals would be held responsible in the same way as if environmental damage were caused, for instance, or human rights laws were knowingly and willfully breached. We have to apply this across the sector.

I am tight on time, so perhaps, in one minute each, the witnesses could outline what measures they have in place and tell us why, as legislators, we should not start to hold individual directors of companies responsible where companies knowingly and willfully allow online harms to be distributed.

I will clarify something. When a company as large as ours, the very large online platforms present, introduces a product or service, we are required by law to do an assessment of it under the Digital Services Act. I cannot speak for the company that is not present today. I understand that is perhaps one of the areas being investigated.

There is a criminal liability provision in the Online Safety and Media Regulation Act. Ireland is one of the only countries in Europe that has a criminal liability provision for senior managers in our companies.

That is only for failure to co-operate. It is not for knowingly allowing online harms to be rolled out.

The standard is any neglect, which is a lower threshold than willful neglect.

Ireland is one of the only countries in Europe that has any neglect standard when it comes to criminal liability for senior managers in a company. There are many products and services to which we could refer. As a business, we have over 50 presets towards young people. Mr. Collard might explain some of those presets that we have.

Very quickly.

In terms of our platform, we make sure that safety is by default on teens' accounts, so if people are under 16, they will not have the ability to direct message. This is not something they can switch on or switch off. We think that is really important for negating contact risk - the risk of grooming. We make sure that young people are not recommended to contact with older people on the platform, so over-18s and under-18s are not recommended to come into contact.

Does Google wish to respond?

Billions of people come to Google to get access to authoritative, credible, useful information, whether it is through search or Gemini, our AI tool, or whether it is on YouTube and so-----

I use it; there is lots of the good stuff. It is the harmful stuff we want to focus on.

Baked in to those products and the reason people use those products - search has been around a long time and people trust it and use it on a daily basis - is that they can trust in them, and they have baseline protections for young people and for all users. I am one of more than 700 people in the company working on child safety. We invest. We have clinical psychologists. We have experts in child development who inform all of our policies and format decisions.

However, Google will pull a product. If something like the problem of the nudification apps or anything like that appears in any Google product, it will pull it straight away.

The approach we like to take is more of a preventative one when-----

If it suddenly arose, Google would pull it straight away.

We test our products and roll out what they need, so if we need to-----

I get that, but if a product ends up out there and on the market, Google will pull it straight away.

We have done that in the past, yes.

For Meta, again, it is a similar issue. The witnesses will obviously be aware of the testimony of Gráinne Seoige and the concerns there in terms of what happens. That involved a Meta product.

On the question of criminal sanctions and sanctions generally, I come back to the point I was making to Deputy Joanna Byrne. What is the metric we are looking for? The metric I think we are looking for is to ensure the safest possible experience for all users online. That is a better metric than who has the harshest penalties or who has the harshest sanctions to put in place.

With respect to getting results, let us say Mr. Ó Broin was the director of a company located by the canal and he knowingly allowed the river to be polluted, he would be prosecuted under criminal law. If people knowingly allow children's and young people's minds to be polluted, and knowingly and wilfully allow a product, and I am not talking about something that comes to light, then there should surely be consequences for the individuals who do that.

The point I am trying to make is that the outcome we are looking for is that the river pollution does not happen and those things do not happen and that there is behavioural change at the ecosystem level. That is what the regulations should be looking for, rather than looking to lock someone up. There are huge sanctions in place already under the DSA-----

I am not looking to lock anybody up. It is a preventative measure. The reason certain people do not pollute rivers is they know they will be personally held liable. Orange is the new black. If somebody knows that they are personally going to be held responsible, they will not do it. The problem I have with fines is that companies write them off as a business cost. That is the reality of the issue. I do not want to see anybody in prison for this. I just want to ensure we have a safe online environment. I am quite impressed by Prime Minister Pedro Sánchez's remarks. His concerns are ones I would share. The companies represented here are trying to outline what they do, but we know there are companies that are not as willing to engage.

If I may, Chair-----

Very briefly. You have ten seconds.

The idea that companies write things off as fines does not reflect the reality that we have huge compliance teams and we take compliance with these regulations incredibly seriously.

We will dig into that. I call Senator Mullen.

Do any of the witnesses have any objection to this country having a law that would restrict access to social media to people aged under 16?

We will start with Google for a change and go from the middle.

We absolutely share the goal around-----

I want a fairly "Yes" or "No" answer, please.

We share the goal around protecting young people, but we do not believe that restrictions for under-16s are the right answer, for two reasons.

We will keep it moving because I want to follow up.

Okay.

We would support something that put the parents in control, so below that parents would have the option to approve apps.

We support evidence-based measures-----

Now, please-----

-----rather than a blunt instrument that is disproportionate or not evidence-based.

I thank Ms. Moss. I suspected as much because what I believe is that the companies here are all in the psychological manipulation business. They depend on advertising. They depend on keeping people's attention and keeping them online and on their platforms. I do not hear the witnesses speaking about the risk of these platforms to formative minds, how their services might impact on people's psychological development or attention span, the question of whether their stuff feeds addictions or whether their persuasive techniques are harmful in the short term or long term. I find it difficult to think that these companies do not prioritise commercial pressures over child well-being. Lots of experts are saying that the kinds of problems they are causing for children cannot be managed or regulated away because it would hit at the core business model they operate, which is keeping people constantly engaged. I just wonder how the witnesses can credibly say that their companies are in the business of promoting safe online experiences for people because safety is a question of "How long is a piece of string?".

Pornography is the single biggest online danger. Our country, admittedly, has failed to legislate. The witnesses could, therefore, justly point to the lack of legislation in this country. What do the witnesses say to the need for strict age verification to make sure? They know that in certain states in the US, the likes of Pornhub have been driven offline because they are civilly liable if young people access porn. It is as simple as that. What do the witnesses say to their organisations being similarly liable if they are the means of children being able to access pornography? Again, I would like quick answers, please.

I am happy to come in on that. On the Senator's first point about access restriction, we are acutely aware that young people have vulnerabilities, and it is our role and responsibility to help protect against those. I worked in the field of child exploitation for 15 years before joining tech, so I understand the importance of it.

I would like a fairly quick answer.

Our goal is not to maximise the time spent. It is about the quality and enriching that time.

I am not hearing anything very clear there. Can I have a response from the other witnesses?

From a teen accounts perspective, 97% of our 13- to 15-year-olds are in that kind of default private setting. It is key that we know the age of that child. That system-by-design approach is really key. That includes time management and curfews overnight in terms of notifications. What that allows the user to do, and the family, including the parents through parental supervision, is actually tailor the experience to their own needs and the needs of their family to make sure they are getting an age-appropriate experience.

If we make it a requirement that children under 16 cannot access AI platforms or AI-supplying platforms, will Google, as an online platform, have an advantage in those circumstances? What do the witnesses say to the need for AI platforms to enable people to have their questions immediately forgotten or to have a means to have their interactions forgotten? As far as I am aware, AI companions remember what people ask them. Again, what do the witnesses say to those two questions? If we regulate such that children under 16 cannot access social media platforms that give access to AI chatbots, would the likes of Google then have a free run at the pitch?

Obviously, that depends on the scope of the measure. That is one of the big questions - what services would be in scope of a measure like that, if there was a restriction for under-16s? Some countries have talked about social media services but, obviously, there are other services that young people use, such as gaming and so on. A big question for legislators, if they are considering that measure, is about what the scope would be.

The Senator might remind me what the second part of his question was.

I am asking about the right to be forgotten using AI.

I am sorry; yes, I was going to mention that. On Google's AI model, Gemini, we have built in the same controls that we have on our other services where users are in control of their data. If people have a Google account, they can click on their profile picture and get access to the data that has been generated by their interaction with the service and choose to delete it. That is-----

Does Mr. Meade agree it should always be the case that somebody using AI, a companion or a buddy, or whatever it is called, should have an immediate, clear and easy ability to have all of that prior history of contact deleted and forgotten?

We think it is good practice to provide controls.

Should it be the law?

It is the default for under-18s.

Should it be required in all cases as a precondition of access to AI?

I do not know if it should be a legal requirement.

Why not?

We would have to give that some thought.

Is a person not entitled to the protection of not having their AI engagement remembered if they do not want it to be remembered?

We may have to look into what law is there at the moment. I know that with GDPR, there is the right to be forgotten, and users have the right to request that their data be deleted. From our point of view, we already provide that so-----

Thank you. Let me ask one last question. We had a lot of understandable anger about the use of AI online chatbots generating child sexual abuse and exploitation material, and also non-consensual material relating to adults. Is it not possible for the companies to ensure, perhaps with help from legislation, that it is simply not possible to use chatbots to generate nudification or pornographic images? One can have safeguards, like filters, that allow for academic or other purposes. The technology would allow the prevention of such imagery being generated, would it not? Is it not also possible to have mandatory watermarking of AI outputs, where providers of generative AI systems embed an identifiable watermark, so that if you create something, it can be traced back to you? What do the witnesses think of such measures?

I think the first one is absolutely right. With Meta AI and AI Studio, they are trained not to comply with the request to create nude images, so that is baked in from the outset. I agree with that first example. I think we should discourage intimate image abuse or nudification, and really think about the impact of that.

What does Mr. Miles mean by "discourage"? Does he mean discourage or prevent through law?

For example, we took legal action against nudification apps last year. There is a very significant number of generative nudification apps because they are freely available on app stores. It is very important that we pursue that to make sure it does not violate our community standards.

I was reading the opening statements last night, and I found them to be very different. Google's is very vague. It has robust policies, but there is very little detail on anything. I am not saying it is the case, but it is almost as if AI wrote it. It is very vague. Meta's is very similar, with a bit more detail. With TikTok's, the detail is actually quite good, which shows the seriousness of how it is taking this meeting. It talks about moderation systems maintaining a high accuracy at 99.2%, file content removed at 91%, the highest ever rate of proactive removal at 99.3% and the highest ever volume of content removed in under 24 hours at 94.8%. That is obviously being increased year on year. Do the other companies have those figures? Can they provide those figures? Why would they not have provided them today when this is exactly what we are talking about?

Certainly. Those figures are all contained in our transparency reports, which are available online. To be honest, it is a challenge when faced with a three-minute deadline to ensure you are covering all of the points and ensuring you are being respectful to the committee.

Your contribution sets the tone of the meeting. It outlines what you believe is the priority to be discussed.

That is why we set out teen accounts and age verification.

Different companies have different priorities as to what they think is important to put out in a public forum here. From my experience, TikTok seems to be a more positive social media outlet than others. I presume all the companies here would view themselves as different from X. What does TikTok do differently than other social media companies? It is not perfect, by any means - no company is - but what does it do that seems to be a bit more positive than other social media companies?

That is a very generous question. I thank the Senator. I am not talking about another company, but in the way in which we develop our recommender system, for example, we have worked to ensure that we are prioritising compliance with local laws, user well-being, minor safety, regional compliance and also content diversification. What we are attempting to do, as a business, is input into our recommender system content that is both relevant to you, but also aligned with our community standards and with the law. We recognise, of course, that there are more than 2.4 million people in Ireland coming to TikTok every single month. Our mission is to inspire creativity and bring joy as a business, but our number one priority at all times is safety.

As a business that is a second generation platform, we are a much younger platform, so we have learned from some of the trial and error of the companies that have come before us. We try to embed safety by design in everything we do as a business. That is reflected in the people we hire. For example, Mr. Collard is formerly of the NSPCC, the UK equivalent of the ISPCC. We have people who come from the United Nations and Amnesty International. Everything we are doing is aligned with our community guidelines, but also with independent external advice. It is critical, as a business, that we are not engaging in groupthink, but engaging with independent external experts, who bring a diversity of opinions, diversity of voices and diversity of experience and lived experience into our business, and that then makes its way onto the platform.

With regard to the moderation of content, the witnesses spoke about how AI plays a role in that, but so do individuals. Meta is also going down the AI route in relation to social media moderating. Is there a fear that, with AI contributing to lay-offs, the knowledge of local regions across the world will be affected with regard to offensive material? Does Meta have enough people working on this?

We have been using AI and machine learning, or whatever version we want to call it, for a very long time. It is set out in our transparency report how that is working on the proactive side to improve the accuracy. I spoke earlier about ensuring that we have the mix right between the AI and the human, and that will continue. It is precisely to the Senator's point about ensuring that we have the right regional nuance when situations arise. I do not see that at any stage it will be just AI that will be moderating content. There will always be a mix between AI and humans.

Do all three companies believe that Grok nudification tools are wrong? Is this actually just the tip of the iceberg? Is it wrong that Grok can do this?

Yes, absolutely.

Yes.

Yes.

If Google believes this is wrong - Mr. Miles mentioned this before I asked the question - why can people still download it on the Google app?

One of the reasons that I say it is wrong is that the material it was producing is definitely not in line with the policies for the Play Store. When that was brought to our attention, we immediately contacted the developer to ensure they created safeguards, and made it clear to them that the app would not be carried on the Play Store.

When did Google do that?

This goes back a week or two ago. I do not have the exact dates. It was at the same time that this was in the media. As the Senator may know, it was the start of this year when this issue came to light.

Was Mr. Meade aware of this issue before it came to light in the media?

No. As I understand it, this was a product change that was undertaken by that company to make this available, or to reduce the guardrails that were available in the app. That is my understanding.

Has Grok replied to Google?

There has been engagement between the teams-----

What has that engagement been?

They made representations that they have changed their policies. They have brought in place new measures to ensure the type of material that was being produced is not being produced.

Does Mr. Meade believe that?

I am not going to say here that this is a completely closed issue, but we are following the process that we have with any app developer. If we see there is a general purpose app-----

When does Mr. Meade expect a decision will be made?

It is an ongoing situation.

Is Google happy that Grok is on its app store, given no change has been made by it at the moment?

The change is being made by the content rating body, I understand, to make the app 18-plus from now. It was previously lower. We see that as a significant step forward. We are still in very regular dialogue with the developer about improvements that can be made.

When I asked if the witnesses agreed with the nudification content of Grok, Mr. Meade was very quick to say "No", but Google is still going to keep that app on the Google store.

Our understanding is that changes have been made to prevent that happening.

That is not the point, but is Meta still happy to keep the app available for people to download?

Only if it complies with our policies. The other thing I should mention is that we do not automatically go to removal of an app like this one that has other purposes. If it was a pure nudification app, the decision would be to take it down straight away, but this is a general purpose AI tool that has other purposes. Our requirement of the developer is to make these changes. It is the developer's responsibility to have safeguards-----

Have other app stores removed Grok from download?

Not that we are aware of. The other thing to mention is that if it were removed from the app store, it would still be on the phones of people who had already downloaded it, so that would not actually make any difference in that case.

That is not a good enough reason.

I want to clarify one thing before Senator Comyn's contribution. Is the content rating body that was mentioned an external or internal body?

It is an external independent body, called IARC. I cannot quite remember what that acronym stands for. Something about independent ratings.

I thank the witnesses for coming here today. The fact that X is not here is disappointing, but not very surprising.

I will keep the questions brief and I hope the answers will also be brief because there is so much to cover. I noted with interest the less-than-enthusiastic response to Senator Mullen's question about the under-16s ban. Australia, of course, banned it. France, Spain, Greece, New Zealand, India, the United Kingdom and certain states of the United States, have either brought in the ban or are in the process of debating it or bringing it in. I hope Ireland will follow suit. Are those countries and governments wrong?

We will start with the representatives from TikTok.

Would support the introduction of a measure that was done at a European level, but any decision that is made is really up to the Government. We believe that whatever side of this debate you land on, I think we are all in agreement that teenagers should be having an age-appropriate experience, whether that is with online media, print or TV.

Does Ms Moss think they are wrong, yes or no?

We do not agree with blunt instruments such as that.

Does Ms Setter think they are wrong?

We absolutely understand and hear the concerns, but we do not think that is the right approach.

Does Meta also believe that this is wrong?

It is probably too early to say. Some of these bans are only just coming into place or being considered now. There is already-----

Given that the representatives are not only lukewarm, but quite definite that they do not believe in this ban, would they be complying with it if it comes into Ireland or will they be looking to offer ways of bypassing it?

No, we would certainly not offer ways of bypassing the law. I think our position is quite clear. We support a digital majority age, but we also support that, under that age, parents should be empowered to decide whether their teens can or cannot access an app. Whether that is 16 or whatever age, we-----

We all know that children will find a way to bypass their parents' parental controls. Please give me an outline of the benefit of under-16s being on social media?

A large part of why we have designed teen accounts in the way we have for ages 13 to 15 and above is to ensure that they have the ability to use social media in a positive way. I will ask Mr. Miles to drill into what that 13-to-15-year-old experience looks like on teen accounts.

What is the benefit? Why should they be there?

There is a huge benefit. All young people have a right to participate online in a healthy way, be it for their interests, connecting with friends or education. It is a plethora of things that they benefit from. By banning them from that, they have no access to that online. They are not developing online development skills in social media. I think that children between the ages of 13 and 16 have the right to participate.

So they have a right to participate. Do people then have a right to target them with child exploitation, child abuse images and grooming?

The reality is we take that kind of content down, and that is not most young people's experience, for the most part. Otherwise, they would not stay on the platforms in the large numbers they do. If we look at the take-up rate on teen accounts, nearly 97% of tens of millions of teens have stayed in the platform. This suggests that they are having a positive, privacy-preserving experience.

I will skip to the representatives of TikTok with this question. Why should a child be allowed to be on social media? What benefit is it?

There are multiple benefits to being online, whether it is to learn, to play or to find their community. There are lots of NGOs, including my last NGO, the NSPCC, that recognises there are inherent benefits. However, there are risks as well and we agree that we need to differentiate between 13-to-15-year-olds, 16-to-17-year-olds and adults who are on the platform and make sure that they have different experiences so that we can avoid the risks the Senator talked about. She mentioned some of the contact risks of grooming. We do not allow under-16s to have direct messaging on our platform so that they cannot be approached. We make sure that they cannot recommend their content in the "for you" feed for other people to see. We make sure that the experience is something that is safe for them.

I want to drill down a little bit further to see how we can increase this protection. It has been brought to my attention that an 11-year-old child is on TikTok at the moment. She is clearly seeing self-harm content because she is sharing this content and commenting on it herself. Who is going to be responsible for that child being on that platform? Who is responsible for taking that child off that platform? How is she seeing that self-harm content? As a 56-year-old woman on TikTok, I would never see anything like that. How come she is seeing it?

In terms of making sure that under-aged people are not on the platform, we take quite a considered approach. First, we try to make sure that they are not on the platform to start off with. This means that if they set up the app store correctly - if their parents set that up - we can stop them from having downloaded the app before then. We have a neutral age gate, which means that when they sign up to be on TikTok, they enter a date of birth. We do not prompt them to say they have to be over 13. We just ask for that. If they say that they are under 13, we immediately stop their ability to create a new account. However, we recognise that we cannot always trust the age that people give us. This is where we start to use AI technology, which we have already spoken about a bit today, to try to estimate the age of users. When we have an estimate that they are younger than the age they have provided us, we send them to human moderators. We also provide the ability for people to report under-age accounts. Be it in the app or on the web browser, if someone has the username of the account being referenced, they can report it and it will be sent to a human moderator who will look to review that account.

How long does that process take? Clearly, this is not an isolated incident. This is a child who can see and share that content, and it is still there. How is that content still there?

We do not permit content that glorifies self-harm. That is prohibited by our terms of service on the platform. There is content that might not be harmful to be seen in isolation, for example, content that references sadness or uses Eeyore the donkey. That content might not be against our terms of service but we recognise that, when seen in combination with other pieces of content, it could cause harm.

That is why I believe a child under 16 years of age should not be there in the first place.

I am going to direct a quick question to Meta. I note that the first line of Meta's statement says, "Meta believes in freedom of expression". I feel that, sometimes, the statement could stop there, certainly judging by some of the content. I cannot read my own social media content because I could not handle it psychologically. I could not get out of bed in the morning if I were to read it all. A child has not got that ability. Has Meta lost control of the platform and its content?

No, certainly not. I am sorry that the Senator is having a negative experience. The experience that politicians, in particular female politicians, have is something we have been quite focused on in Ireland. We worked with the task force on safe participation in public life during the previous Oireachtas. I do not believe we have lost control of the platform. In fact, if we look at something like teen accounts, those are a very positive development in ensuring that the most vulnerable users have the appropriate experience.

Adults are not getting that kind of protection.

Adults are subject to a different level of protection than teens, but I would not accept that we have lost control of the platforms.

Ar dtús báire, tabharfaidh mé deis do Dhualta Ó Broin ceist a fhreagairt trí Ghaeilge, mura miste leis, ar son na n-eagraíochtaí foilseacháin trí mheán na Gaeilge.

Má tá sé sin ceart go leor leis an rannóg aistriúcháin.

Cad iad na beartais bhreise ar féidir le Meta úsáid a bhaint astu chun leanaí a chosaint ar líne?

Ceapaim go bhfuil muid tar éis é a chlúdach ansin. Ní hamháin gurb iad na beartais, ach is é an tslí a bhfuil an tseirbhís curtha ar fáil againn. Le teen accounts, tá sé deartha le cinntiú go bhfuil an t-eispéireas is sábháilte agus is príobháidí acu ar na seirbhísí atá againn.

Tá Meta ag brath ar na tuismitheoirí don mhórchuid, an bhfuil?

Níl. Nuair atá tuiscint againn ar an aois atá ag na daoine óga, táimid in ann iad a chur isteach sna grúpaí seo. Don ghrúpa 13-15 bliana d'aois, níl smacht iomlán ag na tuismitheoirí ar an tslí a n-úsáideann na páistí an tseirbhís, ach tá an-smacht ag na tuismitheoirí air.

Ní féidir leis na páistí é a athrú gan na tuismitheoirí ag teacht chun é sin a dhéanamh. Don ghrúpa níos sine, 16-----

I am going to stick with that and switch to English. I want to talk about parents. All of our guests have mentioned parents. I am not comfortable with the narrative about parents feeling empowered, parent centres, parent guides, family centres and education hubs. With every generation, there is a new worry for parents to deal with. Social media and access to it by pre-teens and teenagers are definitely the worry at the moment for parents. Twenty to 30 years ago, it was probably smoking. If tobacco companies came before an Oireachtas committee and told politicians to just look at the parent guide and they said parents were totally empowered, it would be an absolute farce. We do not have the same level of empowerment over our own teenagers, our own children, when they have a phone. You cannot turn around and just say to a parent of a 14-year-old that they gave them the phone. Many children actually need phones for school at this point. They are being used for certain apps and so on. Therefore, what I have referred to is not realistic.

I am concerned about the narrative about parents having control and being responsible for the insane amount of technology in children’s hands. If you try to take the phone off a 14-year-old to apply certain settings, the best of luck. In real terms, there are ten- and 11-year-olds using TikTok. I do not know how they are doing it, but they are finding loopholes in everything, and that means they are seeing stuff they should not be seeing. I would like to hear some reassurances for parents because they are scrambling. This social media business is a runaway train and I do not think we are doing enough.

I will start, if that is all right, and then let Mr. Miles come in. Listening to parents and understanding what parents are looking for is precisely what we are doing. We are not putting this on parents. Parent guides, parent supports and supports we give to the Parents Council are in addition to all of the regulatory frameworks in place to mitigate risk and ensure we are doing the right thing.

However, the talk of a ban has come up because it feels like we are not doing enough. It keeps coming back to parents. Parents have enough to be dealing with and juggling.

Over the past five years, we have tried to listen to parents rather than tell them what they should be doing.

There are two things. With teen accounts, teens are automatically transferred to stricter settings straight off. They can opt their parents in if they want to change those settings to less strict ones. That does not place any onus on the parents at all but it does involve and create teachable moments for them and the teens, should they want to do it. Teens find that really reassuring.

The other thing is to keep listening to parents and teens in the process. When you release a product, you have got to keep listening. Parents told us that our content policies were a little vague in some areas and that they wanted to be guided by something they could understand. Therefore, we introduced a PG-13 film comparison as guidance for content. Parents said they understood that and related to the content.

Parents are time poor and always very worried about these things, so the onus should not be put on them. They should be allowed to participate when they need to through parental supervision, maybe at download or through teen accounts. In the end, it is about continuing to listen to them and refining products so they will be reassured.

I was alarmed when Ms Moss was asked about her goal for best practice and answered that she wanted to minimise harm to young people. Zero tolerance should be the goal as opposed to minimising harm.

The existence of violations or problematic content on any platform does not equate to permission from us. No large-scale platform of our size can realistically guarantee zero violations.

Unless young people are not online.

The goal is to get as close to zero as possible, but the harsh reality is that it will never be zero. It will never be for as long as there are bad actors attempting to circumvent our processes, moderation and multilayered approach. To give some context, 99% of people have an entertaining or educational time on platforms. The overall percentage of violative content is 0.5%. We are aggressively addressing the 0.5% so the 99% can continue to have a positive experience.

Can you screen grab on the TikTok app?

Yes.

On the Instagram and WhatsApp apps, you can screen grab. Screen grabs are a big problem. If I try to screen grab a work email because I want to tell somebody about it, I cannot do it because it is banned on the phone or app or by my organisation. This is something that could be baked into an app. The testimony Ms Gráinne Seoige gave on the harmful content being spread online when she was involved in a campaign indicated much of it was because material was being forwarded and screen grabbed. One can have all the controls one wants on a teen account, but if somebody screen grabs it and sends it to a teenager, it does not matter what kind of account he or she has. He or she will see everything. Where do the delegates stand on banning or blocking the ability to screen grab and share content on the apps?

An opening statement referred to non-consensual intimate imagery that is not allowed to be generated, but it can be shared. What is the point if it can be shared so easily?

We have a notification control that would not allow you to send a nude image anyway. It would be hashed automatically.

Very briefly.

When you try to transfer it, it will ask, “Do you need to send this image?” The person receiving it would also have it hashed and would have to allow it through. A huge proportion of adults and teens are using this to stop intimate images from being shared. In fact, 45% of people shown the commentary image asking whether they wanted to send the image decided not to. Therefore, there are things that can be done in the design of the product to reduce the virality.

If a person has a swimsuit on, is the image pornographic?

Any kind of nudity at all.

I welcome everybody. Technology is brilliant. I was able to sit at a meeting that clashed with this one and listen to this one on my ear buds. Therefore, I have not missed anything.

Unfortunately, as alluded to by previous speakers, we are in the land of unintended consequences when it comes to breaches. I wish to ask a straight question initially. I am aware that much of the content moderation is probably through AI. How many people do each of the organisations have moderating, watching or supervising the AI?

I would not have the global figure.

Would any of the witnesses?

I do not believe we publish the total. We do have some figures in our Digital Services Act, DSA, transparency report regarding the number of moderators who cover different languages. I do not believe we have the overall figure for the number of moderators.

TikTok has approximately 4,500 moderators for European languages, which is one of the highest numbers among our peers.

How many of them are devoted to English?

It is available in the transparency report. I do not have the number off the top of my head.

Would the other organisations be willing to provide the information afterwards?

I will be very happy to send our DSA transparency report, which will provide a breakdown. A caveat, because this has come up before, is that we are required to report moderators who have moderated content in a particular language. Sometimes, there is just not that much content in a particular language queued for moderation and therefore it does not show up as a large number of people covering that language, but that does not mean we do not have the capability to moderate in that language.

Does “capability” mean AI?

No. We have native language speakers. We also have access to machine translation-----

On moderation, I do not need accurate numbers but could an approximate ratio or figures be provided for the amount of content reported versus the amount detected?

I am very happy to forward the transparency report. Our transparency report is essentially built on determining the proportions. For something like terrorist content, AI is extremely good at catching that.

For something a little nuanced or subjective like bullying or harassment, the human report will be more important.

I would just like to clarify a previous response from Mr. Miles in respect of Facebook and Instagram. I am happy to deal with the WhatsApp element in answer to a separate question.

We publish details of where automated systems are taking decisions on content. A very high level of content in specific verticals is flagged by automated systems before ever being reported by us. As Mr. Ó Broin said, AI is particularly good at certain content verticals. Violent extremism would be one and child sexual abuse imagery would be another. In those categories, we have achieved a very high percentage, namely 95%, or in that area, of content that is flagged by an automated system before it is reported to us on the platform.

I can answer in percentages but I do not have total numbers, if that is still helpful. Some 99% of it is removed proactively by TikTok. Over 80% of it is removed before it even gets a single view. In other words, before anyone even sees it. Over 70% of it is removed within 24 hours. In the context of Digital Services Act reports, TikTok has the highest accuracy rates in moderation among all major social media platforms.

I will move on. Two previous speakers referred to Spain and the steps it proposes to take. Obviously, there were comments regarding a media personality who said that what is proposed is tyrannical and fascist in nature. I do not think anybody here is going to use those adjectives. Do the witnesses believe that taking a step like that would be contrary to freedom of speech?

No. We support a move at European level for the introduction of an age verification requirement. Broadly speaking, we would support it happening at European level. Scope-wise, it should apply to all apps. We support a digital majority age. Underneath that, parents would have the power to approve whether a child has access to an app or not.

Okay.

YouTube does not have traditional social features like many other platforms. It is essentially the world's largest video library. People come to stream music or watch football. People watch it as a streaming service in the living room as a family. It is used in a large majority of schools. We do not have certain measures. For example, we do not have direct messaging for under-18s, we do not have friends lists and we do not connect users by who they know through their social graph. We also do not have ephemeral content like disappearing messages.

We have a lot of things in place that we think are not socially led. Thinking about young people and their rights to access information and education, we believe it is important for them to have this access. We would that say bans are not the right approach for that reason, but we also do not think they would be very likely to work. We are of the view that they would give a false sense of security, and we may well see kids going to less safe platforms which have not had decades' worth of investment, resource and child expertise built into them or the type of baked-in protections we already have on our platform.

I am not sure if TikTok would like to comment.

We have a very similar position here. We understand why there is such a groundswell of support among parents and politicians. We think regulation is the best way of achieving this, making sure we can accurately balance children's or teens' rights to participate in the online world. A risk has been identified by child safety groups, which my colleague pointed out and which is that teens are likely to move to areas of the Internet that are more unsafe and where there moderation technologies are not in place. Also, this likely breaks down the communication barriers between parents and children, which is often a last line of resource.

Relying on regulation is the most appropriate way to do this. We understand that if countries are going to proceed with looking at this type of legislation, it needs to be applied equally to all technology companies on the basis of the risk they could potentially cause to children as opposed to there being a specific list of named platforms.

I thank the witnesses for that. We will move to Deputy Gibney.

I thank everyone for being here. I first want to declare I worked in Google between 2006 and 2014. My time there overlapped with Mr. Meade. I do not believe it overlapped with Ms Setter, but I am not sure. The witnesses cited Google's motto in their opening statement, "To make the world's information universally accessible and useful." At that time, there was also the secondary motto of "Don't be evil", which, sadly, we have to point out, was retired a number of years ago.

Members are referring to the companies by means of different terms. The Cathaoirleach prefers e-commerce platforms and so on. I like the term "big tech" because it gets across the control of power that a very few companies have within both the tech space and broader society, essentially led by broligarchs in the US. That is an important message. If anything, the Grok scandal in recent weeks has sharpened people's attention to the dangers of online harm. I believe nudification is only one of many ways in which big tech and online platforms are inflicting serious harm on our society and on individuals and groups within it.

I am going to try to widen the lens a little and not look at the specifics around any of the functionalities or some of the proposed solutions relating to age verification. I want to start with regulation and lobbying. Many of the witnesses, as heads of public policy, will conduct lobbying on a regular basis. In recent months, the register shows that lobbying is happening from them and their competitors on things like digital simplification, the AI Act and online safety. We know that digital deregulation is being pushed by the Trump Administration with the support of big tech, including some of the companies represented here. For me, there is a real concern among people that the influence the companies hold internationally is paving the way for ordinary people's rights being eroded online.

Have the witnesses ever lobbied in favour of deregulation at European level? In this, I include the digital simplification package which will row back on GDPR and the EU AI Act. I will put that to each of the companies and maybe they can outline, in a nutshell, outline what the objectives of that lobbying are. I will start with TikTok.

I thank Deputy Gibney for her question. Generally, our lobbying involves being supportive of systemic risks.

Has TikTok lobbied on those specific things?

Not on deregulation.

What about the digital omnibus package? Does TikTok not think deregulation and simplification are the same thing?

No. We are supportive of a systemic approach to regulation that looks at systems and processes rather than individual pieces of content.

What about Google?

Not personally, but I think our lobbying returns will show that we have engaged on the digital simplification package. We have been doing that since the Draghi report, which recommended there is scope to make the digital rule book more straightforward.

Sure. We will move on, because I have follow-on questions.

We are very supportive of the European competitiveness programme. We have advocated for it a number of times. In a nutshell, it is not about removing protections for individual rights-----

That is a perspective I do not share.

-----it is about improving implementation.

At the same time, we are hearing that the companies the witnesses represent and other tech firms are lobbying the US Administration to threaten Europe into deregulating online spaces. How do the public affairs and policy arms within their companies interact with such conflicting approaches?

It is probably obvious, but TikTok is not a US-owned business.

We operate in 100 different countries, so we engage with many governments. It is no surprise we would be engaging on both sides of the Atlantic.

Does Mr. Meade see the conflict in those two approaches? The companies involved are saying that one is about simplification and making matters easier for businesses. However, they are also backing Trump to push for greater deregulation in Europe.

Our agenda is fairly consistent. We see the scope for simplification in the digital rule book in the EU, and that is the position we would take with any government.

Okay.

Similarly, it is a very consistent position. It is not just US companies that are calling for simplification. About 60 European companies are also calling for it.

That is fine. The witnesses can understand the discomfort people feel with big tech, when, on one hand, the companies involved are saying that it is not about deregulation but about simplification, the Draghi report or whatever the stock answer is and, on the other, they are telling the Trump Administration across the water to literally bully the EU into rowing back on the digital rights of our citizens. Do the witnesses accept that changing what data counts as private for GDPR purposes, means that it will be legal to keep more sensitive personal data? That is part of the digital omnibus package.

We are probably not across the digital omnibus package enough to answer that question properly for the Deputy.

That is part of it. Do the witnesses accept that self-certification, when it comes to risk under the EU AI Act, will result in weaker protections for citizens and more scope for abuse?

I do not think we can answer that here today.

That is a level of detail I do not have today. I can give a sense of what our position is on the AI Act, if that is helpful.

I will put the same question to Ms Moss.

My position is the same as that of Mr. Ó Broin.

The witnesses do not have enough detail on the digital omnibus package.

What do the witnesses mean when they say that they cannot speak on this level?

The digital omnibus package is not something I have been working on directly, and neither has Ms Setter.

At the beginning of the meeting, I was very clear that the witnesses are here to represent their corporate organisations.

Of course, but we have a topic that we have prepared for. I apologise if this was to be included in the scope of the topic, because that was not our understanding. It is not something we would refuse to engage with the committee on.

I will give Deputy Gibney back her time, but I want to ask about the witnesses not being confident enough to answer this question.

I do not want to mislead the committee by giving answers that are-----

Honesty is the best policy.

Perhaps Mr. Ó Broin would be willing to give a more theoretical answer on the two examples I provided. We know that in the digital omnibus package the recategorisation of data will mean that it will be legal to keep more sensitive personal data will be legal and that self-regulation is, essentially, the shift that is happening with the digital omnibus package in order to allow companies to self-regulate and define what is high risk, which, at present, is a function of the EU AI Act. Do the witnesses agree that this is problematic and is rowing back on users' rights rather than streamlining matters for businesses?

I am happy to follow up with our position on these very specific questions. In terms of the AI Act, at a very high level, when it began and when it was introduced, a risk-based approach was taken. This is something that everyone supported. What happened in recent months is that it moved from being a risk-based instrument to being a technology-specific instrument, and that is the main focus.

It is problematic because it is being framed as harmless streamlining to make it easier for businesses. What we are seeing, however, is a very clear rowing back on people's protections and rights before the AI Act is even implemented, and it is reaching so much further into GDPR.

I want to get into the question of addiction in the short time remaining. After a faulty redaction of internal TikTok documents by the state of Kentucky's attorney general's office, which has subsequently been widely reported on, we know that internal TikTok research revealed that its algorithm was potentially harmful to and addictive for children and adults. DCU research shows that less than an hour after users sign on, harmful content is provided to them. This shows that algorithms seek to make it addictive to users. I appreciate that algorithms are the secret sauce. In this case, the representatives from Google can answer on behalf of YouTube, which is the product that this question most relates to. Do the witnesses agree that algorithms are designed to be addictive and to keep our eyes on the screen at a fundamental level?

No, we would say that algorithms are here to provide users with content experiences that they enjoy, to help them find communities and to access a range of content they might not necessarily have spent time looking at.

Mr. Collard does not think that algorithms are designed to keep our eyes on the screen.

They are not designed to be addictive.

We believe that algorithms help to sort and make practically useful vast amounts of content. There is something like 500 hours of content a second uploaded to YouTube.

Making information universally accessible; not keeping my eyes on the screen.

No, we prioritise user experience over time spent.

Yes, time. I think addiction would be an overly simplistic term. It is about making age-appropriate personalised experiences but also dealing with harms from a preventative perspective and algorithms do a very good job of that.

I respectfully disagree.

The witnesses are all very welcome. I thank them for coming in. How many incidents of online abuse were reported last year in the context of Meta?

In our transparency report, we have a breakdown by harm type or category of community standards, and this is done on a quarterly basis.

All community-----

This is with regard to the various community standards, whether violent extremism, bullying or harassment. They are broken down in the transparency report and I can share them.

I am conscious of time, and am already 20 seconds in. How many incidents of online abuse were reported?

I do not have the full number, but I can share the transparency report with the committee.

How many were referred to the Garda last year?

The way in which we work with the Garda is that it has a law enforcement portal which is open to every member of An Garda Síochána. This is supported by a law enforcement outreach team that supports the Garda when it comes to us looking for information or evidence.

How many were referred to the Garda?