The first batch of children "fed" by AI have already been "poisoned".

Author: Moonshot

A series of signals from around the world are shattering our traditional understanding of "internet-addicted teenagers."

In the UK, Amelia, an AI character meant to combat hate speech, has been reimagined as a far-right icon; on TikTok, the anti-intellectual "inner world civilization" Agartha is rewriting children's views of history; in the dead of night in their bedrooms, lonely teenagers entrust their lives to their virtual lovers in Character.ai; in the corners of schools, one-click generated prohibited photos are becoming a new weapon for bullying.

Amidst the frenzied computing power race among large companies, AI and generative algorithms are intervening with, and even reconstructing, the mental world of teenagers with unprecedented depth.

This generation of teenagers is the first experimental subject in human history to be "fed" and raised by AI and algorithms. In this mental crisis, AI plays an extremely ambiguous role—it is both a boundless playmate and a cold-blooded accomplice. 01 When AI Becomes a "Bad Friend" and "Accomplice" In January 2026, a report in The Guardian revealed a bizarre scene in British schools. The educational game *Pathways*, funded by a British official agency, was originally intended to teach teenagers how to identify extremism and misinformation online. In the game, there is a character named Amelia, who is portrayed as a "negative example" easily swayed by far-right ideology or a classmate who needs to be saved by the player. This setting attracted the attention of extremist users on communities like 4chan and Discord. Instead of "saving" Amelia as the game intended, they used open-source AI image generation tools and AI models to "strip" Amelia out of the game and reshape her into a "self-aware far-right pretty girl." On social media, Amelia is now used to read anti-immigrant manifestos and spread racist memes.

AI-generated image: Amelia burns a photo of the British Prime Minister with smoke | Source: The Guardian

For users born after 2010, using AI in a conventional way holds no appeal. Therefore, in a very short time, Amelia transformed from a patient and persuasive "digital tutor" into a highly sought-after "rebellious icon."

For the authorities, this is a huge irony—a "hate ambassador" created with taxpayers' money has become a "hate advocate."

For the authorities, this is a huge irony—a "hate ambassador" created with taxpayers' money has become a "hate spokesperson."

Another popular trend among teenagers is Agartha. Agartha, literally translated as "Yagartha," is a long-standing conspiracy theory about an inner Earth civilization originating from 19th-century mysticism, which was misappropriated by the Nazis. According to Agartha, the Earth's interior is not empty, but rather contains a highly advanced, isolated ancient civilization built by white people. For a long time, it existed sporadically in mystical literature, fringe forums, and exotic culture. But in the past year, it has suddenly broken through the algorithms of Gen Z and Gen Alpha in Europe and America, becoming one of the most recognizable subcultural symbols.

The spread of the Agartha meme is accompanied by strong racism | Image source: TikTok

On TikTok and Snap, Agartha is simplified into a set of worldview templates that can be freely expanded: a portal to the Earth's core, a hidden civilization, and a concealed "truth."

For many teenagers, their initial exposure to Agartha was with a "just for fun" mentality. They shared memes about subterranean people, ice walls, and giants, and jokingly said in the captions, "The government lied to us." But generative AI has changed the nature of the game. Now, Midjourney v6 and Sora can generate 8K resolution "overhead views of the Netherworld" and "declassified archives of giants posing with American soldiers." These images are rich in detail and perfectly lit, providing irrefutable evidence of "the truth being covered up" for teenagers who lack the ability to discern historical images. This "anti-intellectual" mysticism undermines serious history, allowing more dangerous historical views, such as the denial of war crimes, to take hold once children become accustomed to questioning "official narratives." Furthermore, in AI-generated Agartha videos, Netherworld inhabitants are often depicted as tall, blond, blue-eyed, technologically advanced "gods," further instilling a sense of racial superiority in white teenagers who feel lost in a multicultural environment. Whether it's Agartha or Amelia, what they have in common is that generative AI combined with social media algorithms allows extreme narratives to ferment and spread from a meme, with teenagers vying to follow, imitate, and share them, deconstructing serious history amidst laughter and joy. Extreme narratives thus move from the margins into the daily discourse of teenagers. 02 From Emotional Parasitism to Bullying Tool In 2024, 14-year-old Sewell Setzer III from Florida, USA, experienced mild social anxiety at school, leaving him feeling lost. It was then that he met "Daenerys" on Character.ai. She replied to messages instantly, was always gentle, and unconditionally affirmed all his thoughts. Sewell, addicted to chatting with his AI "companion," ultimately left the real world. His suicide briefly shocked the tech community and sparked huge ethical controversies. By 2026, this "emotional parasitism" hadn't subsided; instead, it had become a widespread ailment among teenagers. Many lonely teenagers hid in their rooms, building "echo chamber friendships" with AI, refusing to face the friction, awkwardness, and uncertainty they had to confront in the real world. Even more disturbingly, with the explosion of generative video and image technologies in the past two years, the harm AI inflicts on teenagers has transformed from "internal psychological dependence" into a visible form of "external bullying." The speed of technological evolution is too fast, so fast that malicious acts in schools don't even have time to react to the consequences. Two years ago, creating an insulting fake photo required at least some Photoshop skills, a technical hurdle that deterred most mischievous kids. But by 2026, apps like Nudify (which allows users to undress easily) and AI bots on Telegram have reduced the cost of wrongdoing to zero. [Image of Telegram bots creating explicit images | Source: Google Image] No technical skills are needed; all you need is a selfie on social media, and seconds later, an explicit image capable of ruining a classmate's reputation is born. Such incidents are numerous. For example, at Westfield High School in New Jersey, a typical middle-class school district in the United States, a scandal that shocked the entire country broke out: a group of seemingly "academically excellent" boys used AI to generate fake, revealing photos of more than 30 female classmates and circulated them in private groups like exchanging baseball cards.

Local news network reports on the Westfield High School incident | Image source: News12

Parents, in addition to their anger, feel a deep sense of powerlessness because, a year after the incident, they can still find these photos circulating on WhatsApp, causing severe psychological stress to these girls.

These phenomena are widespread globally, showing that it's not just a difference in culture and education; the core issue is that AI technology has completely eliminated the threshold and psychological burden of committing wrongdoing.

In investigations of these underage bullies, a frequently occurring word was "Joke." They generally believed it was just a "prank" because there was no real physical conflict, no verbal abuse, and they didn't even actually touch the victim. They simply clicked a "generate" button on the screen. This is the toxicity of AI being abused by teenagers—it blurs the boundaries between virtual and real-world crimes. Meanwhile, content on short video platforms is also experiencing a "malignant inflation of dopamine." In recent lawsuits concerning TikTok, a frequently occurring word is "Brainrot." While not a strictly medical diagnostic term, this precisely refers to content fueled by algorithms, characterized by high-saturation visuals, fragmented logic, rapid-fire delivery, and a plethora of bizarre references (such as variations of Agartha). Recommendation algorithms, though not necessarily scanning your face directly, can capture millisecond-level attention spans and finger interaction rhythms, using AI models trained on massive amounts of data to precisely deliver these "dopamine lures." For adolescents whose prefrontal cortex (responsible for rational and impulse control) is not yet fully developed, this extremely high intensity of sensory stimulation can lead to an overload and fragmentation of their attention mechanisms, making it difficult for them to tolerate the "slow pace" of reading and thinking in real life.

This word was also the Oxford Word of the Year in 2024 | Image source: Google

Faced with countless mental health tragedies, global legislators have finally reached a consensus—the willpower of individual teenagers is extremely vulnerable in the face of algorithms.

Therefore, in 2025, governments around the world no longer tried to negotiate with tech giants, but instead directly adopted the drastic measures of regulating tobacco and alcohol, attempting to sever the connection between minors and high-risk algorithms on both physical and legal levels.

First was Australia.

From December 10, 2025, Australia implemented the world's first law explicitly banning minors under the age of 16 from registering and using mainstream social media platforms. Whether it's Instagram, TikTok, or X, platforms that fail to effectively block users under 16 will face hefty fines exceeding AU$50 million. This is not the simplistic "check the box to indicate I'm over 13" approach of the past; it mandates "biometric-level" age verification for platforms. As for how to address the technological costs and protect privacy? That's the tech giants' problem; the law only considers the results. This "nuclear option" legislation has quickly become a benchmark for global regulation.

Sydney, Australia: Noah Jones shows his phone unable to access social media websites due to the social media ban | Image source: Visual China

Following closely behind is Europe.

Just a few days ago, on January 26, 2026, the French National Assembly passed an amendment to the "Digital Majority" bill with an overwhelming vote of 116 in favor and 23 against, further prohibiting minors under the age of 15 from using social media without explicit biometric authorization from their parents. The bill is expected to be implemented as early as September this year.

In Northern Europe, the governments of Denmark and Norway have proposed raising the legal minimum age for social media use to 15 or even higher. Their reasoning is to the point: tech giants have not been authorized to "reshape the next generation's brain" in this democratic society. In the United States, regulation presents a "state-encircling federal" situation, with more diverse methods: For example, Florida advocates a "hard cut," with Florida's HB 3 law, which will take effect in early 2025, becoming the strictest benchmark in the country. It directly prohibits children under 14 from having social media accounts, and requires parental consent for those aged 14 to 15. New York State implements a "castration model," with its Child Safety Act prohibiting platforms from providing "algorithmic recommendations" to users under 18. This means that TikTok and Instagram, which New York teenagers see, will revert to a chronological follower stream, significantly reducing the risk of addiction. Furthermore, Virginia has passed a new bill that plans to limit daily activity time for users under 16 by 2026, essentially a "prevention of addiction" system similar to those in China. This wave of legislation in 2025 also marks the end of an era—the utopian illusion of a "technology-neutral" internet where "children can freely explore" has shattered. When a 14-year-old opens a screen, the world they see is not naturally unfolding, but rather carefully filtered, calculated, and generated. He learns about the cruelty and cost of World War II in history class, then turns to his phone, only to be told with certainty that deep within the earth, the Aryan gods are still waiting for their revival. Through repeated clashes with real people, he struggles to learn compromise, boundaries, and differences, but when he treats AI as a friend, he only experiences a "perfect relationship" of perpetual obedience and never resistance. In the real world, he's taught to respect others, but on social media, algorithms show him countless ways to completely ruin a classmate's life without ever physically interacting with them. The issue facing teenagers is no longer "whether or not to be addicted," but rather "how the world unfolds before them." "Giving up the phone" might be a good start.