AI and Teenagers: Safety, Mental Health and the Future of Learning

Artificial intelligence is becoming part of everyday learning and communication, raising new questions about how teenagers should use it safely and responsibly.

OpenAI is moving toward a more age appropriate version of ChatGPT for teenagers, reflecting growing concern over how generative AI may affect young people's mental health, judgement and online behaviour.

The company's stated approach is not simply to restrict teenagers from using AI, but to build stronger safeguards around the way they interact with it. OpenAI says it has already introduced age prediction, parental controls, learning features and additional protections for users it believes are under 18. These protections cover areas including self-harm, graphic violence, unhealthy body-image content and sexual or dangerous roleplay. 

The move comes amid a much larger question confronting the technology industry: what happens when increasingly powerful conversational systems become part of a child's everyday emotional and intellectual life?

That question cannot be answered merely by adding a few filters.

Teenagers Are Turning to AI for More Than Answers

For teenagers, ChatGPT can be a tutor, writing assistant, brainstorming partner and source of information. It can explain a difficult mathematical concept, help structure an essay or generate practice questions.

Used properly, this can be enormously valuable.

AI is increasingly finding its way into classrooms, raising questions about how technology can support learning without replacing it.

OpenAI itself has increasingly emphasised learning features designed to encourage students to understand a problem rather than simply receive an answer. Its teen-safety approach also includes tools intended to support deeper learning and critical thinking.

But the same conversational quality that makes AI useful can create risks.

Unlike a search engine, a chatbot responds conversationally. A young person can return repeatedly to the same system, discuss personal problems and begin treating the machine as something closer to a companion than a software application.

That is where the mental health debate becomes particularly important.

As AI becomes increasingly conversational, teenagers may begin to see chatbots as more than tools for information and learning.

OpenAI says it has worked with more than 170 mental health experts to improve how ChatGPT recognises distress, responds to sensitive conversations and directs people towards real-world support when appropriate. Its safety work specifically addresses self-harm, suicide, severe mental-health symptoms and emotional reliance on AI. 

These safeguards are important. But they should not create the impression that an AI system can replace a parent, teacher, counsellor, psychologist or trusted adult.

A chatbot can respond to distress. It cannot assume responsibility for a child's wellbeing.

The critical-thinking problem

There is another concern that is less dramatic but potentially more widespread.

The challenge for educators is ensuring that AI supports students' learning without replacing the thinking, questioning and problem-solving that make learning meaningful.

Education is not simply about obtaining correct answers. Children learn by struggling with difficult questions, making mistakes, revising their reasoning and eventually reaching conclusions themselves.

If AI routinely completes that intellectual work, students may become dependent on answers without developing the ability to arrive at them independently.

This is why the emphasis on critical thinking is significant.

The better educational use of AI is not:

“Write my assignment.”

It is:

“Challenge my argument. What have I missed?” The difference may determine whether AI becomes a shortcut or a learning tool.

Schools therefore need to teach students AI literacy, not merely AI usage. Children should understand that AI can be inaccurate, biased or incomplete and that an apparently confident answer is not necessarily a correct one.

Parents cannot outsource responsibility to technology companies

Parental involvement remains important as teenagers navigate AI, social media and other digital tools.

OpenAI's parental controls allow parents to link their accounts with a teenager's account and customise certain settings. The company has also introduced controls around features such as voice mode and image generation, alongside notifications for certain high-risk situations.  These tools are useful, but parental supervision cannot be replaced by software settings.

Parents need to know how their children are using AI, not merely how much time they spend using it. A teenager using AI to understand physics is very different from one using it as their primary source of emotional support.

The solution therefore requires a combination of technological safeguards, family involvement, school-level digital literacy and access to qualified mental-health support.

The Indian question

India's growing use of AI in education makes questions of digital literacy, access and responsible use particularly urgent.

This debate has particular relevance for India.

The country's young population is rapidly adopting AI, while its education system is simultaneously attempting to integrate artificial intelligence into classrooms. That creates an opportunity, but also a responsibility.

Indian schools should not wait for AI-related problems to emerge before establishing rules.

Students should be taught:

• how to verify AI-generated information;

• when AI assistance constitutes academic dishonesty;

• how to protect personal information;

• why emotional dependence on AI can be problematic;

• how algorithms can reproduce bias; and

• when a human teacher, parent or professional should replace an AI conversation.

The digital divide must also be considered. If responsible AI education is available mainly in elite schools, AI could widen existing educational inequalities.

Safeguards are necessary but transparency matters too

OpenAI's move is welcome, but it also raises difficult questions. How accurately can an AI system determine someone's age? What happens when an adult is incorrectly treated as a teenager or a teenager is incorrectly treated as an adult?

How are safety interventions evaluated? How often do safeguards fail? And how should parents understand what information is being used to make these decisions? OpenAI says its services are not intended for children under 13, while users under 18 require parental or guardian permission under its current privacy policy. Such rules need to be accompanied by clear explanations that ordinary families can understand.

The larger lesson

The instinct to protect teenagers from the risks of AI should not turn into an instinct to protect them from AI itself.

Young people will enter universities and workplaces where artificial intelligence is increasingly unavoidable. Shielding them completely would leave them less prepared for the world they are about to enter. The objective should instead be responsible exposure. Teenagers should learn to use AI without surrendering their judgement to it.

They should be able to ask a chatbot for an explanation without assuming it is an authority. They should be able to use it for learning without allowing it to do all their thinking. And they should understand that when the subject is deeply personal, particularly mental health, an AI system should be a bridge towards human support not its replacement.

OpenAI's new safeguards are therefore an important acknowledgement of a problem that the technology industry cannot ignore. But the responsibility cannot rest with OpenAI alone.

Parents must remain parents. Teachers must remain teachers. Mental health professionals must remain professionals. And children must remain the people making the decisions about their own lives not passive recipients of decisions made by machines. 

The real test of AI safety will not be whether a chatbot can refuse the most dangerous question. It will be whether technology can help young people become more capable of thinking, questioning and connecting with the real world rather than less.