The new homework: Teaching children to question AI answers

The arising challenge, according to pundits, is that the same systems can produce convincing but false information, fabricate sources and generate manipulated images, leaving children to distinguish useful assistance from answers that only sound correct.

Cybersecurity firm Kaspersky, in a September 3 advisory, says children need to learn to verify AI-generated information, recognise so-called hallucinations and avoid sharing personal or school-related information with AI services.

Kaspersky’s observations show that children’s interest in AI tools continues to grow “as these technologies become more accessible and integrated into everyday learning.”

“In fact, AI is likely to become as commonly part of their studies as search engines, online dictionaries and educational videos.”

The company’s guide comes as international education agencies shift their focus from whether children should encounter AI to how schools, parents and students should manage its use.

Unesco’s AI Competency Framework for Students recommends teaching children to develop human-centred attitudes, understand AI ethics, acquire technical knowledge and eventually participate in designing AI systems.

The framework places these competencies across three stages—understand, apply and create—reflecting a move toward preparing students to work with AI rather than treating the technology solely as a threat to academic integrity.

This reflects the modern-day reality that the availability of generative AI changes what it means to complete an assignment independently, particularly where a student can obtain a polished response without demonstrating how they reached it.

A child asking a chatbot to write an essay may receive a coherent answer within seconds, but the speed of producing the response can remove the research, reasoning, and writing practice that the assignment was designed to develop.

According to Kaspersky, learners need to start treating AI as an assistant that can, among other things, explain concepts, suggest arguments, identify weaknesses in a draft, or generate practice questions rather than completing schoolwork for the student.

A student using AI to solve a mathematics problem, for example, should be able to explain the method independently and reproduce the solution rather than simply transferring the chatbot’s response into an exercise book.

The same principle applies to research since a fluent AI response does not establish that its underlying information is accurate, current, or drawn from a genuine source.

“When a child can generate an essay or receive a finished answer to a math problem in seconds, it may be tempting to submit the result without understanding it. This can save time in the moment, but it prevents the child from developing the very skills the assignment is intended to practice,” it says.

“Parents can agree with their children that AI may help explain a concept, suggest a structure, provide examples or ask practice questions, but it should not complete the entire task on their behalf.”

Unesco has warned that generative AI can create fabricated information and that education systems need safeguards because the technology is advancing faster than many regulatory and institutional responses.

A global Unesco survey of more than 450 schools and universities found in 2023 that fewer than 10 percent had formal guidance covering generative AI, highlighting how quickly the technology had entered education ahead of institutional rules.

In Kenya, the issue gathers particular relevance as the government’s National AI Strategy 2025-2030 identifies education among sectors where AI and digital skills are being integrated into the country’s broader technology agenda.

The strategy’s implementation roadmap also identifies limited access to devices, gaps in teacher training, and weaknesses in data privacy and security as challenges that could constrain digital education.

This means AI literacy cannot be reduced to teaching children how to write better prompts because they also need to understand the limits of the systems producing the responses.

Generative AI does not independently establish truth before producing an answer, meaning an apparently authoritative explanation can contain a wrong date, invented quotation, non-existent study or flawed reasoning.

Unicef says children are increasingly turning to AI chatbots for information, learning and creativity, while evidence on the effects of the technology on their cognitive, social and emotional development remains limited.

The privacy question becomes even more complicated when children begin using AI conversationally, as a homework prompt can easily contain private details about the student, such as their school, classmates or family.

A child asking for help with an assignment might paste an entire school document, upload a photograph of a worksheet, or include names and personal circumstances without considering the information as sensitive.

Kenya’s Data Protection Act requires parental or guardian consent before personal data relating to a child is processed and requires processing to protect and advance the child’s rights and best interests.

The Office of the Data Protection Commissioner has separately told the education sector that minors cannot provide valid consent on their own and that schools must ensure appropriate safeguards when processing children’s information.

For schools adopting AI tools, this puts data governance alongside academic considerations, requiring institutions to understand what information platforms collect, why it is processed and how long it is retained.

Unesco’s guidance, similarly, calls for privacy protection and age-appropriate approaches to the use of generative AI in education, while urging institutions to assess whether particular tools are ethically suitable.

→ [email protected]

Follow our WhatsApp channelfor the latest business and markets updates.