Revealed: The TikTok safety loopholes exposing children to vile rape and torture cartoons
An investigation by The Mail on Sunday has found that TikTok's "restricted mode", a safety setting used by millions of parents to shield children from harmful content, is failing to block graphic AI-generated videos depicting rape, torture and violence. The footage exploits loopholes in TikTok's rules, which prohibit "realistic violence" and "sexually suggestive content" but do not adequately cover animated depictions, allowing companies to profit by pushing disturbing clips designed to be addictive to underage users.
The investigation identified start-up Director AI, run by former Snapchat director Ben Chiang, as producing tools behind more than 15,000 such videos across 50 accounts, with one clip alone viewed over 120 million times; Chiang has said the videos have amassed over 10 billion views and openly called the output "AI slop". Testers who set up a restricted TikTok account for a 14-year-old found their feed still flooded with violent and disturbing material, while a companion app, Reel AI, rated suitable for children as young as seven, charges per episode after free previews, including one showing a mother electrocuting a baby. Shadow Education Secretary Laura Trott condemned the failure, and psychotherapist Dr Catherine Knibbs accused the firms of "creating wealth at the expense of children", while Chiang argued Director AI merely provides tools and is not responsible for how creators use or publish them.
- TikTok's "restricted mode" fails to block graphic AI-generated violent videos.
- Director AI's tools produced 15,000+ clips viewed billions of times.
- Politicians and experts demand accountability from platforms and creators.
New here? Start with this
Children exposed to violent AI-generated content on TikTok despite a safety setting called "restricted mode," which many parents rely on to filter out unsuitable material. The concern centres on animated clips depicting rape, torture and other extreme violence, which have been able to circumvent TikTok's usual restrictions because the platform's rules on "realistic violence" and "sexually suggestive content" were not written with AI-generated cartoons in mind.
At the centre of this is Director AI, a start-up founded by Ben Chiang, a former Snapchat director, which makes tools used to create this kind of content. A related app called Reel AI, aimed at users as young as seven, has also been highlighted for hosting disturbing material behind a paywall. Both are commercial ventures that generate revenue from content viewed by very large audiences, including children.
This matters because it raises questions about how well social media companies protect young users from harmful material, and about the responsibilities of AI tool-makers whose products can be used to create such content at scale. It also touches on wider debates in Britain about online safety regulation and how existing rules keep pace with rapidly evolving AI technology.