TikTok's AI algorithms expose children to harmful and sexually explicit content
TikTok's recommendation algorithm reportedly pushes sexually explicit content and search suggestions to users as young as 13 despite age restrictions and content filters being enabled. Amnesty International and Global Witness investigations found the platform's search and For You feed algorithms actively promote inappropriate material to minors and content related to self-harm to users showing signs of psychological distress.
What happened
Global Witness conducted an investigation in October 2025 using seven TikTok accounts set to age 13 with Restricted Mode enabled. The organisation found that TikTok's search algorithm actively pushed highly sexually explicit search suggestions and content to these accounts, with some receiving explicit recommendations on their first click in the search bar.
Amnesty International Taiwan reported findings from 2023 showing that when test accounts demonstrated signs of psychological distress on TikTok's For You feed, the algorithm rapidly promoted large volumes of content related to depression and self-harm. When Amnesty International contacted TikTok in 2025 to ask whether the issue had been addressed, the platform only referenced existing features without proposing substantive reforms.
According to Taiwan data, 46.1% of children aged 9–12 use TikTok, potentially exposing approximately 380,000 students in grades 3–6 to risks of privacy violation and exposure to age-inappropriate content. TikTok's stated age verification mechanisms function only nominally; 42.3% of Taiwanese teenagers aged 15–18 report using TikTok multiple times daily.
Reported impact
Exposure to sexually explicit content and search suggestions violates TikTok's own community standards, which prohibit any content involving minors that is nude, depicts sexual acts, or contains sexual suggestions. The violation of this stated policy despite the existence of age restrictions and content filters raises concerns about moderation failures.
Research by Amnesty International and mental health professionals indicates that prolonged exposure to inappropriate content, combined with the platform's algorithm design that prioritises engagement, poses risks to psychological health and creates potential links between online experiences and self-harm behaviour. Taiwan data show correlations between exposure to intimate image distribution and youth self-injury risk.
Index analysis
TikTok's age restrictions and content filtering mechanisms (13-year minimum age requirement, Restricted Mode) failed to prevent the delivery of sexually explicit content and search suggestions to the youngest permitted users. The search algorithm's proactive promotion of explicit content to restricted accounts indicates the recommendation system operates independently of stated content policies.
The platform's failure to implement effective algorithmic safeguards despite prior warnings from international human rights organisations (Amnesty International's 2023 findings) suggests a pattern of inadequate response to documented harms. TikTok's 2025 response focused on existing features rather than algorithmic changes, indicating limited substantive modifications to address the identified risks.
Governance controls
TikTok stated that upon learning of Global Witness's findings, it immediately launched an investigation, removed policy-violating content, and improved search suggestion functions. The platform claimed to remove 90% of policy-violating videos before they are viewed and stated it deletes approximately 600,000 underage accounts monthly through technical detection and staff training.
Taiwan's Democratic Experiment laboratory (2025) found that content lacks effective review and age-appropriate classification mechanisms on TikTok. Taiwan legislators have proposed amendments to artificial intelligence, child protection, and personal data protection laws to establish age verification, mandatory platform accountability mechanisms, and algorithmic transparency requirements.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.