partnered with the AI Elections Accord to set expectations for how to manage the risks arising from deceptive AI content. In partnership with the Coalition for Content Provenance and Authenticity (C2PA), TikTok enhanced its auto-labelling by incorporating content credentials that attach metadata to content, enabling the instant recognition and labelling of AI- generated material. TikTok also joined the Adobe- led Content Authenticity Initiative to collaborate with a cross-industry ecosystem focused on restoring trust and transparency online. This capability has been implemented for images and videos, with development still under way for audio-only content. To help TikTok’s community understand how to use its tools, the platform launched a video campaign focused on explaining the origins of content on TikTok. Developed in consultation with the international human rights organization WITNESS, the campaign also encourages users to report content that appears AI-generated but lacks proper labelling, raising awareness of the issue. TikTok partnered with MediaWise, a programme of the Poynter Institute, to create videos focused on universal media literacy skills, explaining how tools like AIGC labels can help provide additional context for content. Additionally, TikTok is collaborating with industry peers to support the National Association for Media Literacy Education’s new AI Literacy Initiative, aimed at spreading AI literacy to a wider audience. TikTok also draws essential external advice from its network of Safety Advisory Councils, which include nine regional councils and a US Content Advisory Council. These councils bring together experts from diverse fields, such as youth safety, free expression and hate speech. Their insights play a key role in shaping policies, refining product features and helping TikTok stay proactive in addressing emerging safety challenges. Implementation To implement its counter-misinformation strategy, TikTok combines technology and human expertise to combat misinformation at scale. This effort includes specialized misinformation moderators equipped with advanced tools and training, as well as local teams that collaborate with experts to ensure responses consider regional context and nuances. TikTok has also partnered with 19 global fact-checking organizations that assess content accuracy in over 50 languages and help enforce its misinformation policies. TikTok’s endorsement of the International Foundation for Electoral System’s Voluntary Guidelines for Election Integrity for Technology Companies is crucial, given the key role platforms play in communicating election information. These guidelines set expectations and practices for companies and election authorities to promote election integrity. By implementing measures to provide trustworthy information to users, TikTok helps maintain public trust in its platform during critical election periods. TikTok’s Edited Media and AIGC policy mandates that creators label AIGC or edited media that features realistic-looking scenes or individuals. This can be accomplished using the AIGC label or by including a clear caption, watermark or sticker. Through the partnership with C2PA, TikTok automatically labels AIGC uploaded from certain other platforms. Additionally, TikTok’s policies prohibit content that misrepresents authoritative sources or crisis events, as well as content that falsely portrays public figures in contexts such as being bullied or making endorsements. The platform also disallows the use of likenesses of young people or adult private figures without their consent. In addition to labelling, implementing a series on media literacy content addressing topics such as misinformation, AIGC and AI transparency helps prevent confusion among viewers who may not understand the meaning behind these labels and also helps them to detect misinformation more easily. Feedback, measurement and transparency There are several metrics that offer insight into TikTok’s AI and digital literacy efforts. The first is focused on the auto-labelling tool. Between September 2023 and May 2024, 37 million creators have used the tool that auto-labels AIGC made with TikTok AI effects. Relatedly, as the first video-sharing platform to implement content credentials, the increase in auto-labelled AIGC on TikTok may be gradual at first. As other platforms also implement content credentials, TikTok will be able to label more content. TikTok collects and publicly shares data related to community guidelines enforcement. The data is granular, and relevant sub- metrics can be helpful in understanding how TikTok’s policies are put into practice. This data includes: –TikTok’s proactive removal rate for civic and election integrity, misinformation and synthetic and manipulated media was nearly 99%, and over 89% of those videos were removed before any views. –Between the week of 7 July 2024 and mid- September 2024, TikTok has removed over 250,000 edited pieces of media and AIGC that violated their policies. Since September 2023, 37 million creators have used the tool that auto-labels AIGC made with TikTok AI effects. The Intervention Journey: A Roadmap to Effective Digital Safety Measures 19

The Intervention Journey A Roadmap to Effective Digital Safety Measures 2025