The Intervention Journey A Roadmap to Effective Digital Safety Measures 2025

Example technical interventions TABLE 1 Intervention Description Example organization(s) AI-powered content labelling systemsAI-powered systems that classify content into different categories (e.g. misinformation, hate speech, explicit material), allowing platforms to flag, rate or limit the distribution of harmful content.YouTube (Google) Uses AI to label content as safe or harmful, influencing recommendations and the visibility of videos. Real-time AI and human-in-the-loop moderation systemsHybrid content moderation systems where AI flags content and human moderators make the final decision, improving moderation accuracy for nuanced or context- sensitive content.Twitch Identifies potential violations of community guidelines using a combination of automated detection and user reporting. Nudity detection systemsAI-driven technology harnessing on-device machine learning to detect and blur (when turned on) nudity images. Functions in secure, encrypted environments and may default to active protection for younger users, with prompts encouraging broader adoption among adults.Instagram (Meta platforms) Leverages AI/ML built into the app on user devices. When turned on, it automatically blurs images containing nudity that are being sent and/or received in Instagram direct messages. Teens have it turned on by default, adults need to opt into the feature. Directing people to safety tips when sending or receiving these images, developed with guidance from experts about the potential risks involved. User identity verificationTechnologies to verify users’ identities using methods such as facial recognition, ID verification or behavioural analytics.Tinder and Hinge Uses identity verification tools to ensure users are real, reducing catfishing and online abuse. Blacklisted URL systems for illegal contentSystems that block access to known URLs containing illegal content such as CSAM, preventing sharing or viewing of such content.IWF Works with internet service providers and tech companies to block URLs hosting CSAM. Hashing technology for illegal content detectionTechnologies that use hashing algorithms to detect and prevent the spread of illegal content, such as CSAM or terrorism-related media.PhotoDNA (Microsoft) A hash-matching technology that can be used to detect previously identified CSEA material. Source: YouTube Official Blog. (2023). Our approach to responsible AI innovation; Twitch. (n.d.). Safety at Twitch; Instagram. (2024). New Tools to Help Protect Against Sextortion and Intimate Image Abuse; Match Group. (n.d.). Safety; Internet Watch Foundation. (n.d.). URL List; Microsoft. (n.d.). PhotoDNA. 3.2 Educational interventions Educational interventions in digital safety consist of structured programmes, activities or other informational resources designed to inform and educate individuals and groups about digital risks, safe online practices and cybersecurity measures. The goal of these interventions is to enhance users’ awareness, knowledge and skills for navigating the digital world safely. Educational interventions are crucial in raising awareness and providing targeted knowledge. These interventions must build on existing knowledge, progressing through different levels – similar to a school curriculum – rather than starting from scratch. Capacity building for educators is equally important, requiring investments in training and resources to equip them to handle sensitive topics. Furthermore, harnessing technology such as large language models (LLMs) can enhance the learning experience by offering safe, simulated environments where individuals can explore scenarios without real-world risk, facilitating practical and secure learning opportunities. The Intervention Journey: A Roadmap to Effective Digital Safety Measures 30