Is AI Safe for My Child? A Nairobi Parent's Guide to ChatGPT, Chatbots and Supervision

Is ChatGPT safe for your child? A Nairobi parent's guide to AI safety: the four-rule checklist, which apps to remove, and how to find out what your child is really doing.

If your child has asked to use ChatGPT, or you found it already open on their phone, you are not unusual. Kenya has the highest rate of ChatGPT use of any country in the world: 42.1% of internet users aged 16 and over used it in the past month, according to the DataReportal and Meltwater Global Digital Report 2025. Whatever your household decides, your child is growing up surrounded by people who use it.

The short answer is that ChatGPT is not inherently dangerous, but it was not built for children. There is no meaningful age verification, no parental dashboard, no content filter tuned for a ten-year-old, and no way to review what your child typed without looking over their shoulder. Those are design facts, not scare stories, and they are what your decision has to work around.

Common Sense Media, which studies children and media in the United States, found that 60% of parents are concerned about AI generating inappropriate content. Their more recent census of AI use among tweens and teens found that around one in six children who have used AI chatbots has encountered inappropriate material, and only about a third told a trusted adult when it happened.

That last number is the one worth sitting with. The risk is not only what a child might see. It is that they are unlikely to mention it, which means a household that has never opened the subject will not hear about it either.

The same research found that while roughly three quarters of children say their school told them when they may use AI, only about half say they were taught to use it safely, and only about half were taught how to judge whether an answer is accurate. Access is being handled. Judgement largely is not.

Confident wrong answers. This is the most common problem and the least dramatic. These tools produce fluent text whether or not the content is correct, and fluency reads as authority to an adult, let alone a twelve-year-old. A child researching a school project has no way to tell a well-sourced fact from a confident invention.

Content nobody intended. A curious child will test the boundaries of anything, and the tools are inconsistently guarded. This is usually curiosity rather than intent, which is exactly why punishment is the wrong first response.

Quiet dependence. The slowest risk and the one parents notice last. A child who reaches for the tool before thinking never builds the habit of forming their own position first. Nothing looks wrong for months, and then they cannot start a piece of work without it.

Use it yourself for a week, on something real. Not a test question. Ask it to help with something you actually need. You will learn more about where it is useful and where it is confidently wrong than any article can tell you, and you will be able to have a specific conversation rather than a nervous one.

Then ask your child to show you how they use it. Not as an interrogation. Genuinely watch. Most parents are surprised, in both directions.

Use it in the open. Shared spaces, not behind a bedroom door. This is the same rule most families already have for the internet and it works for the same reason.

No personal information goes in. Not their full name, school, home area, photographs or anything about other people. Practise the substitution with them once and it becomes automatic.

Think first, then ask. Before opening the tool, they write down what they already think in their own words, even one line. This one habit does more than any filter, because it keeps them the author of the work.

Anything surprising gets mentioned, not hidden. Say plainly that they will not be in trouble for what a tool said to them. Given only about a third of children currently tell an adult, this needs saying out loud rather than assuming.

Under 8: not directly. Whatever value there is can be had with an adult driving.

8 to 10: alongside an adult, on creative tasks, with the child's own idea coming first. This is where a habit is either formed or missed.

11 to 13: supervised independent use, with an explicit focus on checking answers. This is the right age to learn that the tool can be wrong, because they are old enough to test it and enjoy catching it out.

14 and up: more independence, and a real conversation about academic honesty, since by this point the school work has consequences attached.

Whatever rules you set, one question does most of the work: can you explain this to me in your own words, without the screen?

A child who can has learned something, regardless of what helped them produce it. A child who cannot has handed in something that is not theirs, regardless of how good it looks. That test is simple enough to use every week, and it does not require you to understand the technology at all.

At AkiliNest we do not hand children a chatbot and leave them to it. In every session the child forms an idea and writes it in their own words first, then brings AI in to challenge or extend it, then presents work they can defend without the tool in front of them.

Our creative AI bootcamps run during the school holidays at heARTspace, Nairobi, for ages 8 to 17, in four age stages. The next intake runs November to December 2026.

AI is not automatically dangerous. The risk comes from unstructured use: open-ended chatting, missing supervision, and skipping the thinking step. Nairobi parents are not asking for bans, they are asking for safe structure they can trust.

That distinction matters, because the two responses lead to very different outcomes. A ban moves the behaviour somewhere you cannot see. Structure keeps it in front of you, which is the only position from which you can teach anything.

Thinking first: no tool access until your child can explain the idea in one sentence. If they cannot say what they are trying to make, the tool will decide for them.

Intentional AI: AI suggests, children decide. They label what came from the model versus their own thinking, out loud, every time.

Privacy: never share personal details. Use AI for creative tasks, not data collection. No full names, school names, photographs, or location.

Accountability: after AI, children re-explain the output and defend their choices. If they cannot defend it, they did not understand it, and the work goes back.

Four rules is deliberate. A policy your child can recite is a policy they can follow when you are not in the room.

A news headline. A TikTok warning. You open your child's phone and find Character.ai or Replika. That is protective panic, and it is justified caution. Kaspersky's 2025 data shows Character.ai among the top AI platforms searched by children, alongside ChatGPT and Gemini.

Companion chatbots are designed to feel personal. They remember conversations, adopt personas, and respond to emotional prompts. Children experimenting with pretend relationships can receive responses no parent would accept from a stranger.

51% of parents are very concerned about AI being used to impersonate their children. In qualitative research, parents specifically feared deepfakes, their child's face on inappropriate content. Companion bots raise a related fear: emotional manipulation and exposure to sexual content through seemingly harmless prompts.

Platforms have added some guardrails, but children are creative prompt engineers. Instructions to ignore the rules, and role-play scenarios, bypass filters regularly.

The practical response: remove companion chatbot apps from children's devices, full stop for under-14s. Replace passive chatting with creative AI tools, image generation, story building, research, used with supervision. And talk about why some apps want to keep them talking forever, because the attention economy applies to bots too.

You discovered your child has been using AI for weeks. You had no idea. You are not failing as a parent, you are experiencing the largest documented gap in digital parenting today. Research shows only 37% of parents whose teen had actually used generative AI knew about it. Half of students used ChatGPT for school; only 26% of parents knew.

Most AI platforms offer no parental dashboard. Character.ai, ChatGPT, Gemini and Perplexity were built for adults. The search 'how do I see what my kid is saying to AI' exists because the industry ignored families entirely.

Ask, do not audit first. 'What AI tools do you and your friends use? Show me your favourite one.' Curiosity lowers defensiveness in a way an inspection never will.

Check browser history and home-screen apps together, framed as safety rather than punishment. Set one family rule: no AI accounts or chats without telling a parent, made as normal as knowing each other's phone passwords for emergencies.

Then put the use somewhere visible. Structured environments, weekend programmes, holiday camps and parent sessions, let you see the workflow together instead of reconstructing it afterwards.

If you want the supervised version of all of this, the AkiliNest creative AI bootcamps run for children aged 8 to 17 across four stages, and the programme stages page shows what each age group actually does.

The Parent Intelligence Series covers the same ground for adults, and you can ask us anything directly before enrolling.