Children, AI Chatbots and the Safety Gap

When Elena Sinel — founder and chief executive of Teens in AI — appeared on Sky News on 12 June 2026, the presenters asked what was happening in children's bedrooms and classrooms as AI becomes part of daily life. Her answer was direct: "It is worrying times at the moment."

The interview focused on a statistic that surprised the presenters: around seven in ten teenagers now say they use AI every day. For Elena, the number itself is only part of the problem. The bigger issue is that many young people are using AI without being taught what it is, how it works, what it can get wrong, or how it may affect them emotionally.

AI is becoming more than a homework tool

One of Elena's strongest warnings was about emotional dependency. Teenagers are not only using AI to answer homework questions. Some are using chatbots as places to talk, ask for advice, share worries, and disclose personal experiences they may not tell an adult.

That creates a serious safeguarding concern. A chatbot can sound calm, warm and understanding. But it does not understand a child. It does not know them. It is not a friend, a teacher, a counsellor or a trusted adult. It is a system generating responses.

That difference may be obvious to adults working in AI. It is less obvious to a teenager who feels lonely, anxious or unsure where else to turn. Elena warned that some young people are already forming bonds with chatbot systems. They may trust them with secrets, follow their suggestions, or return to them repeatedly for comfort.

She also pointed to documented cases where prolonged, unguarded interaction with chatbots has been linked to delusion, psychosis and emotional dependency. This is not safe by default, and it should not be treated as harmless experimentation. Children are using tools built by adults, released quickly, and made available before schools, parents and regulators have had time to deal properly with the risks.

Safeguarding needs to apply to chatbot companies too

Elena was not arguing that young people should be kept away from AI. Her point was that children need protection, education and clear boundaries before these tools become normalised in their lives.

The government has started to talk more seriously about holding technology companies accountable. Elena welcomed that. But she argued that accountability must include the companies building the chatbot systems children are now using.

If a young person turns to a chatbot while distressed, the system's response becomes a safeguarding issue. How it replies matters. What it encourages matters. Whether it recognises risk matters. Whether it keeps a child talking when they need human help matters. At the moment, many of those choices sit with the companies designing the tools — and that is not good enough when children are involved. These are exactly the design and accountability questions covered in our guide to what ethical AI means in practice.

Schools are being asked to cope without enough support

The interview also turned to schools. Almost half of schools say they are not prepared for AI — how to use it, how to detect it, how to teach about it, and how to protect pupils from the risks. Teens in AI has explored this in its white paper, The AI Readiness Gap Starts Early, which has been submitted to the UK government.

Elena highlighted three gaps:

• Teacher training is inconsistent. Many teachers still think of AI mainly as ChatGPT, when AI is much broader. Teachers cannot guide pupils properly if they have not been trained themselves.
• School policies vary widely. Some independent schools have started putting guidance in place. Many state schools have not. Schools need a clear national framework they can adapt, rather than each one having to start from scratch.
• Too much responsibility is left with individual schools. Government guidance exists, but guidance alone does not solve the problem. Schools need practical support, training and model policies.

Assessment has to change

AI also raises a difficult question for education. If pupils can use AI to produce homework, summarise texts, write essays and generate polished answers, then schools need to think carefully about what they are assessing.

Elena argued that the education system has not moved quickly enough. Much of it is still built around forms of homework and assessment designed before pupils had access to these tools. This does not mean banning AI from schools, and it does not mean rushing to adopt every new product. It means asking more practical questions:

• What should pupils be able to do without AI?
• Where can AI genuinely support learning?
• Where does it weaken thinking?
• How do schools assess judgement, understanding and originality when a fluent answer can be produced in seconds?

Elena was also cautious about schools assuming that buying enterprise AI tools automatically makes them prepared. A product such as Microsoft Copilot may be useful in the right context, but buying software is not the same as having a clear educational strategy. Schools need to know what problem they are trying to solve, what pupil data is being used, and whether the tool improves learning.

Young people need AI education now

The interview made one thing clear: AI is already part of young people's lives, and the adults around them are still catching up. Teachers need training. Schools need workable policies. Parents need clearer information. Government needs to take more responsibility. Companies building child-facing AI tools need stronger oversight.

This is where Teens in AI's work sits. Young people should be able to understand AI, question it, use it safely, and recognise when a tool is not acting in their best interests. The technology is already in their hands. The education and safeguarding response now needs to match the scale of the risk.

If you are planning an event or programme on AI safety, ethics or education, get in touch or explore Elena's speaking topics.