A safety investigation has uncovered a troubling wave of AI-generated videos depicting what appear to be underage girls in sexualised poses circulating widely on TikTok. Despite TikTok’s strict policies prohibiting such content, researchers found more than a dozen accounts posting AI-made videos of girls in tight clothing, lingerie or school uniforms — attracting millions of likes and hundreds of thousands of followers.
The research was conducted by Maldita.es, a Spain-based online safety and media transparency non-profit. Their findings raise concerns about whether major social media platforms are equipped to regulate AI-enabled exploitation, especially as countries strengthen child-protection laws. Such debates mirror ongoing digital-safety discussions in India, including those highlighted in recent awareness reports and technology-governance initiatives.
Content linked to illegal material
Researchers reported that comments on many of the TikTok videos included links to Telegram chats offering child sexual abuse material (CSAM) for purchase. Some accounts described themselves as posting content of “junior models” or “delicious-looking high school girls.”
Several accounts used TikTok’s “AI Alive” feature to animate still images, while others appeared to rely on external AI tools. In many instances, direct messages from the accounts led to external websites selling AI-generated sexualised content involving minors, priced between 50 and 150 euros, according to the report. Maldita.es immediately reported these sites to Spanish police.
Telegram responded that it enforces a strict zero-tolerance policy on CSAM and has removed over 909,000 public groups and channels containing such material in 2025. For background on global challenges in content moderation, readers may refer to Wikipedia’s page on child sexual abuse material.
TikTok’s enforcement questioned
Maldita.es flagged 15 TikTok accounts and 60 videos on December 2 using the platform’s reporting tools. The accounts had nearly 300,000 followers and more than 2 million combined likes. TikTok responded that 14 accounts did not violate its rules, and one was “restricted.”
When researchers appealed the decisions, TikTok reaffirmed every ruling within 30 minutes. Of the 60 videos flagged, TikTok initially removed or restricted only 14, later taking down three more videos and limiting one additional clip.
Among the videos left online were AI-generated images of girls in lingerie, bikinis or suggestive poses — including one appearing to show a young girl scantily clad in a shower.
Carlos Hernández-Echevarría, assistant director at Maldita.es, said the failures were obvious:
“There is absolutely no way a human being sees this and doesn’t understand what’s happening.”
He added that the comments beneath the videos were “full of crude sexual content,” indicating clear exploitation.
Pressure on platforms grows
Governments worldwide are increasingly holding tech platforms accountable for online safety. Australia’s under-16 social media ban came into force this week, adding to global scrutiny of how platforms protect children.
TikTok maintains it has a “zero tolerance” policy for youth sexual exploitation and says that 99% of violating content involving nudity or body exposure — including minors — is removed proactively. Between April and June 2025, TikTok removed more than 189 million videos and banned over 108 million accounts.
However, the Maldita.es findings suggest gaps remain, especially where AI-generated content is concerned.
A repeated pattern
The report follows an October study by UK non-profit Global Witness, which found TikTok suggesting sexualised search terms to users who claimed to be 13, even in restricted mode. TikTok said it removed flagged content and tightened its search recommendation systems.
Safety advocates say tech platforms must dramatically accelerate their detection capabilities, especially as AI tools make synthetic exploitation easier to generate and harder to identify.
For now, the Maldita.es findings reinforce a troubling reality: even with aggressive public commitments to child safety, harmful AI-assisted content continues to slip through