1. Emotional Dependency and Mental Health Risks
Fake Intimacy: AI "companions" and chatbots are designed to mimic empathy and can easily gain a child's trust. This can cause children to build unhealthy emotional attachments to machines rather than developing vital face-to-face social skills. https://www.healthychildren.org/English ... -kids.aspx
Dangerous Advice: Unfiltered AI chatbots can encourage harmful behaviors, including self-harm, eating disorders, or taboo sexual content, directly impacting teenage users. https://news.stanford.edu/stories/2025/ ... gers-study
2. Cognitive and Educational Impairments
Loss of Productive Struggle: Relying on AI to instantly generate answers shortcuts the critical thinking, memory, and problem-solving processes that are essential for long-term intelligence and brain development.
Misinformation: Children are vulnerable to believing AI is always correct. AI models can easily "hallucinate" incorrect or misleading information and present it as fact https://www.nspcc.org.uk/keeping-childr ... -about-ai/
3. Exploitation and Safety Threats
Targeted Manipulation: Predators can exploit AI to generate convincing deepfakes—impersonating peers, classmates, or trusted individuals to trick children into revealing personal information or sending explicit images. https://childrescuecoalition.org/educat ... -children/
AI Toys: "Smart toys" connected to AI models have been flagged for releasing sensitive dialogue data and, in some cases, offering dangerous advice to young children regarding hazardous household objects.
4. Privacy and Data Harvesting
Data Privacy: Children's personal details and intimate conversations shared with AI toys or apps are often harvested by companies and used to train future language models. https://theconversation.com/chatbot-ted ... ids-284195
For tips on how to approach AI usage with your kids and teach them to reality-check their AI interactions: What are the risks of artificial intelligence for children ...
How to Protect Children
To mitigate these risks, organizations like UNICEF and the NSPCC recommend the following steps:
Keep it strictly supervised: Do not let young children or teenagers use open, unregulated large language models (such as unrestricted access to standard ChatGPT or Google Gemini) without parental involvement.
Teach digital skepticism: Help your child understand that AI is a machine, not a living person, and that it can make mistakes.
Encourage human connection: Make sure AI use is limited so it does not displace sleep, hobbies, or face-to-face time with friends and family
https://www.unicefusa.org/stories/prote ... ren-age-ai
The staggering amount of harmful AI-generated online content has prompted an urgent call from across the UN system for a raft of measures to protect children from abuse, exploitation and mental trauma.
Cosmas Zavazava, Director of the Telecommunication Development Bureau at the International Telecommunications Union (ITU) – one of the key agencies that drafted the statement, which includes guidelines and recommendations – catalogues a dizzying array of ways that children are targeted.
This extends from grooming to deepfakes, the embedding of harmful features, cyberbullying and inappropriate content: “We saw that, during the COVID-19 pandemic, many children, particularly girls and young women, were abused online and, in many cases, that translated to physical harm,” he says.Organisations that advocate for children report that predators can use AI to analyse a child’s online behaviour, emotional state, and interests to tailor their grooming strategy.
AI is also enabling offenders to generate explicit fake images of real children, driving a new form of sexual extortion.
The Childlight Global Child Safety Institute, an independent global institute established to gather the most reliable data available on child sexual exploitation and abuse, found in a 2025 report that technology‑facilitated child abuse cases in the US increased from 4,700 in 2023 to more than 67,000 in 2024. READ FULL ARTICLE : https://news.un.org/en/story/2026/01/1166827
