Last month, the clip my niece replayed more than any other on her tablet followed a sentient watermelon navigating a love triangle. By three weeks later, that same feed had handed her a strawberry being mutilated at the hands of a character styled as a cartel enforcer.
A new analysis from the Global Network on Extremism and Technology argues that exact slide — from goofy brainrot into something far nastier — is now happening at scale.
The gore is the part nobody’s watching
According to GNET’s researchers, the tide of AI-generated fruit videos washing through social feeds doubles as a distribution channel for quietly seeding “violent and extremist content.”
Taken individually, most of the reels register as noise. Brightly colored fruit characters — pornographic ones, in some cases — act out melodrama: love triangles, getting fired, brotherhood. Ordinary algorithmic filler.
What the analysts documented, though, was a growing share of those reels tipping over into a category they label “AI fruit/vegetable gore”: extreme scenes of torture, mutilation and other cruel acts of violence.
Cartels and far-right coding, in fruit form
One detail sets this apart from run-of-the-mill shock content: the brutality is targeted rather than generic.
The authors note that plenty of these reels frame their violent characters as members of particular groups or adherents of particular ideologies — criminal cartels, far-right groups, extreme nihilism. In other words, the politics is written into the character design itself.
“Violence-based AI slops reinforce a fascination with violence for violence’s sake, thereby allowing the so-called online ‘ode to violence’ to progressively gain ground among young people within the digital space,” the analysts write.
Their argument extends further still: they contend the material can “lay the basis for the consumption of further violence-based audiovisual material, making young users fall into the trap of online radicalization and audiovisual violence consumption.”
Read the report with a hand on your wallet
Time for the caveats, of which this report has earned a few.
The question of whether AI genuinely nudges children toward violence still awaits answers from academic studies. GNET’s report stays mostly at the surface. It even manages to cite at least one AI-generated editorial along the way — an awkward stumble for a document sounding the alarm on synthetic media.
Then there’s who produced it. GNET serves as the research arm of the Global Internet Forum to Counter Terrorism, an outfit bankrolled and staffed by tech industry representatives, and one that has drawn criticism over its part in pushing digital censorship. When an industry-funded body raises alarms about content the industry itself makes, a second read is warranted.
The radicalization angle is the smaller problem
Park the extremism debate for a moment — it’s not really where the weight sits.
Bot farms are pumping this material out at industrial volume. Everyone gets pulled along, and children get pulled deepest, because their media literacy hasn’t finished forming.
The result is a self-feeding nightmare of algorithmically-driven slop. Tuning it to celebrate violence takes no more effort than tuning it to celebrate electoral fraud, racism, or whatever other social ill you’d like to name.
That’s the real takeaway. The danger isn’t that fruit gore manufactures child extremists; it’s that any machine able to produce endless fruit gore on demand can just as readily produce endless anything else, pointed at whoever the algorithm has decided is watching. Swap out one form of depravity and another fits the slot. What makes gen-AI slop dangerous, full stop, is the boundlessness sitting behind it.
The practical lesson here isn’t found in a parental control menu. It’s that the feed has no bottom — and nothing in the way things currently work is laying one down.