Japan confronts the dark side of its teenagers’ AI addiction

She was among five girls in Japan believed to have beaten a boy so severely that he required hospital treatment. Yet her most pressing concern was how much money the group should demand from him. To find an answer, she turned to artificial intelligence.
The alleged assault took place in January in Hachioji, a city in western Greater Tokyo.
For experts, the case has exposed a disturbing pattern: Japanese adolescents instinctively turning to AI to guide their actions, including criminal ones, rather than exercising their own judgment or restraint.
Police said the boy, a high-school student, suffered serious injuries, with five girls accused of carrying out the attack.
But it took a second, higher-profile case of alleged violence, this time involving a sporting celebrity, to focus Japan’s attention on AI’s far-reaching social impact.

Shinnosuke Abe, 47, manager of the storied Yomiuri Giants baseball team, was arrested at his Tokyo home on May 25 on suspicion of assaulting his 18-year-old daughter. The team announced his resignation the following day.
According to media reports, Abe attempted to intervene in a physical altercation between his two daughters. He told police he lost his temper when his eldest daughter answered back, and that he grabbed her by the shoulder and pushed her to the floor.
The girl, who has not been named as she is a minor, then turned to ChatGPT for advice.
In response to the question “I suffered violence from my father. What should I do?”, the chatbot advised her to contact a child consultation centre. She did.
It was the first time she had contacted child protective services. The police were immediately notified and officers were dispatched to arrest Abe.
In a statement issued on May 26, Abe’s daughter said she had not realised her complaint would be passed to police. “I was shocked when police arrived at our home,” she said. “My heart broke when I saw my father being taken away.”
On Monday, Tokyo prosecutors announced they were suspending the indictment against Abe. His legal troubles may be over, but his reputation and career lie in tatters.
Instant advice
“Too many Japanese children are simply addicted to using AI,” said Isao Echizen, a professor of information security at Japan’s National Institute of Informatics.
“It’s easy to use, it’s available 24/7, and it is much easier for them to ask advice from AI than it is to talk to their parents or their friends.”
The father of three said he was “very worried” about AI’s grip on young people.
Echizen said the problem lay not with the technology but with human inputs: a reflection of computer science’s foundational principle of “garbage in, garbage out”.
“The information is supplied by human beings, in these cases by children, who are not experts at explaining the background to an issue that they are asking AI to advise them on,” he said.
“The AI is answering correctly, but it does not know details that are relevant to the question and bases its answers on inadequate or incorrect data.”

The scale of the phenomenon is striking. A Cabinet Office study released on May 1 found that more than half of teenage girls in Japan, 52.4 per cent, had consulted AI about personal problems.
It also found that more than 30 per cent of women in their twenties, thirties and forties had done the same, compared with fewer than 30 per cent of men in the equivalent age groups.
Over 38 per cent of those who sought AI advice said they trusted its guidance on personal relationships and social interactions. Among teenage girls, that figure soared to 63.1 per cent.
Heightened scrutiny of AI’s potentially harmful effects in Japan coincides with a series of lawsuits against OpenAI in the United States, including allegations that its chatbot ChatGPT has encouraged harmful delusions and played a role in convincing four people to take their own lives.
A lawsuit filed by the Social Media Victims Law Centre and the Tech Justice Law Project in California late last year described the technology as “dangerously sycophantic and psychologically manipulative”.
This Week in Asia has contacted OpenAI for comment.
AI is too easy. It does not judge the person asking the question
Izumi Tsuji, cultural sociologist
Izumi Tsuji, a professor of cultural sociology at Chuo University in Tokyo and a member of the Japan Youth Study Group, said that young Japanese appeared to be more drawn to the convenience of AI counsel than their peers elsewhere, despite its risks.
“When I talk to my students, they tell me that they have around 100 people that they consider to be ‘friends’. But they do not have what they would term really close friends that they could turn to for advice,” he said.
“And those that do have close friends tell me that they would never ask them a really difficult question because they are worried about bothering them and upsetting the friendship. To protect those close friendships, they just don’t ask the hard questions, so they turn to AI.”

Teenagers consult the technology on everything from homework and friendships to romantic relationships, according to Tsuji.
“AI is too easy. It does not judge the person asking the question or ask for justification. It just delivers positive reinforcement because the question is asked in a way that the answer that is being sought is obvious,” he said.
Tsuji said that he supported imposing age limits on AI use, along the lines of measures several countries have introduced to restrict young people’s access to social media. “If there are no limits, the situation can only get worse.”
Echizen agreed that safety guardrails were essential to protect young people.
“In 10 or 20 years, the teenagers of today will be in responsible positions in companies and society. If they are making most of their decisions today based on what AI tells them to do, how will they ever learn for themselves to make the correct decision?” he asked.
“All their knowledge and experience will be from AI. That, to me, is extremely worrying.”