AI-generated videos are becoming increasingly common on platforms such as TikTok and YouTube. Many are colourful, repetitive and seemingly harmless, but new research warns that some may expose children to disturbing narratives, violent imagery and even extremist themes.

A recent analysis by the Global Network on Extremism and Technology (GNET) examined the rise of AI-generated animated videos featuring anthropomorphic fruits and vegetables. According to the researchers, this content can blur the line between meaningless entertainment and material that normalises aggression, glorifies violence or echoes extremist propaganda.

The findings do not establish that these videos directly cause cognitive decline or violent behaviour. However, they raise important questions about content moderation, recommendation algorithms and the safeguards required for AI-generated media targeting young audiences.

What is AI “slop”?

The term “AI slop” describes mass-produced, low-quality content created primarily with generative AI tools. It usually involves limited human input and is optimised for views, clicks, engagement and advertising revenue rather than originality or educational value.

Because these videos can be produced quickly and cheaply, a single creator may generate hundreds of clips and distribute them across several platforms. Recommendation systems that reward frequent publishing and high engagement can then amplify the content, helping it reach millions of viewers.

Children are particularly vulnerable to this model. Bright colours, rapid editing, repetitive sounds and absurd narratives are designed to capture attention and encourage continuous scrolling.

Researchers and commentators sometimes connect excessive exposure to this type of overstimulating, intentionally meaningless media with the popular concept of “brain rot”. While this is not a clinical diagnosis, it describes concerns about reduced concentration, compulsive consumption and dependence on instant stimulation.

Harmless characters, disturbing narratives

One prominent category identified in the analysis features animated fruits and vegetables placed in dramatic storylines involving romance, betrayal, domestic conflict and revenge.

Despite their child-friendly appearance, these videos may include aggression, emotional abuse, torture, mutilation, assassinations and mass-casualty scenarios. The contrast between familiar animated characters and graphic behaviour can make the content particularly difficult for children and parents to assess.

Repeated exposure may also contribute to desensitisation by presenting extreme violence as humorous, ordinary or entertaining. The risk is not limited to explicitly graphic scenes. Emotional intensity, escalating conflict and unpredictable storytelling can also create an overstimulating viewing experience.

How extremist themes can enter children’s feeds

GNET’s analysis also suggests that AI-generated entertainment may be used to conceal or spread extremist narratives.

Some videos reportedly borrow language, symbols and storytelling patterns associated with terrorist organisations, criminal groups or violent ideological movements. These references can be presented through fictional characters and viral formats, allowing the material to appear absurd or comedic rather than overtly propagandistic.

For example, a recurring character named “Carrot Khalifah” has reportedly been portrayed telling an army that “the Carrot State will rise again”, language that appears to imitate propaganda associated with the Islamic State.

This does not mean that every strange or violent animated video is designed to radicalise its audience. It does, however, demonstrate how generative AI can make harmful narratives easier to produce, disguise and distribute at scale.

Why platform safeguards struggle to keep up

Platforms such as TikTok and YouTube already have policies restricting graphic violence, terrorism-related material and content that may harm minors. The problem is that AI-generated media can be difficult to classify.

A video might use cartoon characters and playful audio while depicting a deeply violent narrative. Individual clips may appear harmless when reviewed in isolation, even though the wider series repeatedly reinforces aggression or extremist ideas.

The scale of production creates another challenge. Moderation systems must evaluate enormous volumes of content, while creators can rapidly change characters, prompts, visual styles and accounts to avoid detection.

Generative tools have lowered the technical and financial barriers to video production. Yet the ability to create more content has not been matched by equally accessible systems for verifying its origin, assessing its context and managing its distribution.

The broader technology challenge

The rise of AI slop illustrates a wider principle: generating content is becoming easier, but governing it remains difficult.

Safer digital platforms will require more than automated detection. They need age-appropriate product design, transparent recommendation systems, effective parental controls, clearer labelling of synthetic media and consistent human oversight.

Technology companies must also consider how their engagement models influence what gets created. When platforms reward volume and attention without adequately accounting for quality or risk, low-cost AI content can quickly dominate users’ feeds.

For organisations building AI-powered products, safety cannot be added only after deployment. It must be considered throughout the system, from data and model selection to testing, monitoring, escalation procedures and human review.

AI can accelerate creation at an unprecedented scale. The responsibility now is to ensure that control, accountability and user protection evolve just as quickly.

Source