Ms Claire Dil?:

I thank the Chair and members of the committee for the invitation to attend today’s meeting on the topic of the protection of children in the use of AI. I am director of government affairs for Europe at X and I am joined today by my colleague, Ms Niamh McDade, our head of government affairs for Ireland and the UK.

As we aim to build a global town square and provide everyone with the ability to connect, debate and share information, we are committed to ensuring a safe environment for all our users. X’s purpose is to serve the public conversation, and we believe that freedom of expression and platform safety can and must coexist. We welcome the opportunity today to discuss X's work to keep users, especially young people, safe on the platform.

I will start by saying that X is not the platform of choice for children and teens and we do not have a line of business dedicated to children. Users on X must be aged 13 and if a person tells us they are under 13, they will not be able to sign up for an account. According to our data, in the first three months of 2024, users aged between 13 and 17 accounted for less than 1% of X’s active account holders in Ireland. Although minors represent a small fraction of X's user base, we are fully committed to the protection of this group, which is a more vulnerable audience online, and have a number of tools and policies to protect them on our service. X’s age assurance process combines self-declaration of age with additional technical measures to ensure that the account holder’s age is genuine and that appropriate controls are in place to protect children. By default, 13- to 17-year olds have high privacy, safety and security settings in place on their accounts. For example, they will not see sensitive media, including graphic and adult content, and their direct messages are closed and location is turned off. Additionally, advertisers cannot choose to target this age group.

We believe verifying users’ ages and soliciting parental consent for app downloads could play a pivotal role in addressing age verification. This approach could leverage existing processes, filtering all inappropriate apps for minors. It would also act as a privacy enhancer across the ecosystem by avoiding the need for personal information sharing at the individual app level.

X is also a proud member of several child protection initiatives, such as the Tech Coalition, WeProtect, the Internet Watch Foundation and the Children Online Protection Lab, and we continue to welcome opportunities for co-operation with child protection NGOs.

Regarding X rules, we remain steadfast in our commitment to keeping everyone on X safe. Our rules require users to ensure the content they upload and their behaviour complies with our rules as well as all applicable laws and regulations. We also confirm that AI-generated content is subject to X rules and we enforce policies irrespective of the source of creation or generation of such content. We take the opportunity of this hearing to confirm to the committee that X remains committed to the fulfilment of the Digital Services Act compliance obligation and intends to fully comply with relevant legislation in relation to AI.

A number of our policies are particularly relevant to the protection of children, and our supporting statement that we sent prior to the hearing provides further details on our rules. In particular, we confirm that X has zero tolerance toward any material that features or promotes child sexual exploitation. If child sexual exploitation content is posted on X, we simply remove it. Fighting this kind of content on our service is our number one priority as a company, and our policy covers media, text and illustrated but also computer and AI generated images.

We have strengthened our enforcement with more tools and technology to prevent bad actors from distributing, searching for or engaging with CSE content across all forms of media. We also remove any account that engages with CSE content, whether it is real or computer generated. Our priority is that we are able to catch it and take action regardless of whether it has been generated using AI. In parallel, we continue to invest in human and automated protection and content moderation.

Turning to AI-generated content, at the outset, we want to clarify that X does not have a generative AI product live in the EU or Ireland at this time. With respect to misleading media, we have developed and continue to expand important resources. For example, our synthetic and manipulated media policy prohibits users from sharing synthetic, manipulated or out-of-context media that may deceive or confuse people and lead to harm. Furthermore, our community notes product addresses a wide range of sophisticated media types, including AI-generated content, by allowing contributors from a diverse group of people on X to write notes on media wherein they can write a note on a specific image or video and the note will be shown automatically on other posts with matching media. This community-led approach has significantly increased the scale and speed by which potentially misleading media is detected and labelled on the platform. There are currently more than 100,000 active contributors to community notes in EU countries, which accounts for 35% of the global contributor base. To demonstrate the impact of this product, in the last month in the EU region, there have been 130 million note impressions on community notes.

Finally, we come to our recommendation algorithm. First, it is important to note that on X we give users a clear choice over their use of recommendations. People have two options to view posts in their timeline, either "for you” or “following”. Under the ‘‘following’’ tab, they will only see posts from accounts they follow, and under the ‘‘for you’’ tab, they will see recommended posts for them from both their networks. Every day, people come to X to keep up with what is happening and the ‘‘for you’’ tab aims to deliver them the best of what is happening in the world right now. This requires a recommendation algorithm to distil the millions of daily posts on X down to a handful of top posts that ultimately show up in the ‘‘for you’’ timeline to make it easier and faster for users to find content and accounts relevant to their interests. Recommendations may amplify content, so it is important they are surfaced in a responsible way. Our recommender systems are designed to exclude harmful and violating content by integrating with visibility filtering systems, and we have several ways of preventing potentially harmful or offensive content and accounts from being amplified, including using machine learning technology and reviewing user reports.

To increase transparency and accountability, X open sourced its algorithm on GitHub in March 2023. Therefore, it is for everyone to be consulted there, ahead of the DSM coming into force and X’s engineering team published a Blogpost on our website to explain to the public how this algorithm works in a simple way. We welcome feedback on recommendations from people using X receive. For the For You timeline recommendations, feedback can be provided by selecting Not interested in this post/topic. We use this as a signal to recommend less of this type of content to the user.

Additionally, controls are important both in helping people on X to curate their own experiences and to provide good feedback for our recommendation system. A variety of options are available for people using X to control what they do and do not see on our service. Features include, but are not limited to mute and block features, the option to filter notification, extensive privacy and safety settings, as well as the ability to turn off autoplay for video.

We submitted a longer supporting statement before today's meeting which members will have received. We have much more to tell about our policies and rules in regard to this topic but in the interests of time I will stop now. I thank members for the opportunity to appear before the committee today. I look forward to the discussion.