AI videos of fruit and vegetable characters unleashing gory violence are spreading extremist ideas and helping normalise graphic violence, researchers claim.
The content is said to be part of a rise in "brain rot" clips showing anthropomorphic foods playing out melodramatic romance, betrayal and scandal plots.
However, there is a sinister side with "extreme mutilation, torture, and hyper-realistic violence" on show in some videos.
Other clips apparently reference specific criminal and extremist groups such as Islamic State (IS).
The Global Network on Extremism and Technology (GNET) - a research initiative backed by the Global Internet Forum to Counter Terrorism - issued the warning in a report.
It said videos were getting around Tik Tok's rule against violent content as they may appear innocuous at first glance.
They also often used a "cute" or "Pixar-like" appearance to "soften" the image.
One example appeared to replicate content from IS videos with a carrot making a potato "explode", while another featured a carrot "assassinating" a potato with a hammer, the study found.
Another features a "Carrot Khalifah" telling his army "the Carrot State will rise again" - which GNET said alluded to an IS slogan.
Two posts included hashtags often used by supporters of the terror group to disseminate propaganda, the report said.
Gruesome drug cartel killings are apparently mimicked in other clips.
'Potato supremacy'
Other videos are said to reference neo-Nazi organisation Atomwaffen Division and nihilistic subculture No Lives Matter.
Statements and comments pointing to far-right ideology appeared alongside the content, researchers said, with "potato supremacy" said to refer to white supremacy.
"Food violence-based material empowered by AI is easily accessible, putting minors and youth at risk of inadvertently consuming graphic or disturbing content," researchers claimed.
Read more from Sky News:
Meta to alert parents if teens ask AI about suicide or self-harm
What to know about the solar eclipse
They said such content helped normalise "extremist and violent narratives and imagery" and users could also stumble on terror material and head down a dangerous path.
"This type of content can lay the basis for the consumption of further violence-based audiovisual material, making young users fall into the trap of online radicalisation and audiovisual violence consumption," the authors state.
GNET wants TikTok to develop "countermeasures" to limit such AI-generated content. TikTok hasn't yet responded to the report.
(c) Sky News 2026: Violent AI clips of fruit and veg characters could help radicalise young people, researchers
Dogs can differentiate between happy and sad faces, new study suggests
Flight cancelled as child refuses to sit down and fasten seatbelt
Drivers sue Mercedes-Benz and allege car's logo gave them burns
Tarona Taafaki: Boxer lands Tuvalu's first Commonwealth Games medal without winning a fight
