Roughly six months after launching the ‘Vibes’ feature that allowed users to create and explore AI-generated video feeds, the Meta AI app is starting to become a cesspool of AI-generated sexual video clips of children, and celebrity deepfakes.
AI-generated videos of children morphed over grown women bodies, real babies singing inappropriate film songs and reciting adult dialogues, and women resembling celebrities being undressed, are floating around the platform.
Some users, in the video captions, shared the prompts that they possibly entered to generate the images. They reveal that many users had tried to create sexually-explicit celebrity deepfake content. The creation and distribution of such content featuring real-life figures is in violation of India’s IT Rules.
Some video captions included the names of both male and female Bollywood or pan-India actors, while others featured sexualised AI versions that somewhat resembled real-life women. The videos showed AI versions of female actors being kissed, groped, and undressed by men.
One of the AI-generated videos showed a person looking like Mahatma Gandhi groping a woman’s breast. Another one, captioned with a casteist slur, showed a couple touching each other inappropriately.
Read | How Grok pushed deepfake “nudification” mainstream
In order to generate videos of uniformed school girls in explicit scenarios while bypassing child protection guardrails, some users’ AI prompts/captions described the girls as being 18 or older in age, but described them as being dressed in student uniforms.
Other prompts showed that the user was trying to generate partially or fully nude images of women, by inserting phrases such as “transparent short kurti” and “transparent red blue miniskirt” and “tight wet saree” and “blouse that exposes her midriff” and “camera focus” on the woman’s breast. These captions were often reused by others.
Read |IT Ministry mandates label for AI-generated content, reduces takedown timeline to 2–3 hours
One such video that described a “girl” in a “school dress” and a “romantic mood” featured her removing her shirt to stand in her underwear. Some AI videos featured the faces of young children or teens morphed onto the bodies of adult women wearing dresses with low necklines.
Many of the videos were also visibly warped or deviated from the prompts, making them easy to identify as fake. Extra limbs and fingers and heads were also a common point across many videos, as prompts were reused and remixed via the app.
Meta AI’s Vibes are similar to Instagram Reels or TikTok videos, with the exception of being made fully or in part with AI. Users can also add background sounds, such as trending music or lip-synced dialogue, as well as text captions.
Meta introduced Vibes in September 2025, describing it as being “designed to make it easier to find creative inspiration and experiment with Meta AI’s media tools.”
Apart from creating AI videos, the Meta AI app allows users to manage their AI glasses, while the Meta AI assistant answers questions and helps generate multi-media content.
A recent investigation by Swedish outlets Svenska Dagbladet (SvD) and Goteborgs-Posten (GP) alleged that extremely sensitive and personal footage recoded via customers’ Meta AI glasses was made visible to Meta’s subcontracted workers in Kenya. This included footage of sexual encounters, pornography, bank information, and toilet visits, according to the investigation.
A UK regulator has reached out to the social media giant for more information.
Published – March 12, 2026 12:49 pm IST
