As ChatGPT, Claude and Gemini become part of everyday learning, schools face a bigger challenge than detecting AI use: teaching students how to use it responsibly without outsourcing the thinking education is meant to develop.

Generative AI has created one of the most consequential questions facing schools since the arrival of the internet. Students can now ask a chatbot to explain a difficult concept, solve a maths problem, translate a paragraph or draft an essay within seconds. That convenience can support learning, but it can also remove the very struggle through which learning takes place.

In a recent LinkedIn article, Chandrakumar R Pillai, AI Advisor, Enterprise Architect and Co-Founder & President of Gate2ASI, argues that schools should move beyond the question of whether students should use AI at all. His central point is more practical: educators need to teach students when AI is appropriate, where its limits lie and when using it undermines the purpose of an assignment.

That shift matters because AI literacy is quickly becoming less about knowing how to prompt a model and more about knowing when not to delegate human judgement.

A Traffic-Light Model for Classroom AI

One of the most useful examples highlighted by Pillai comes from Cheshire Academy in Connecticut, where teachers use a traffic-light system to clarify the role AI can play in individual assignments.

A green assignment permits AI use. A red assignment requires students to work without it because the teacher wants to assess their own knowledge, reasoning or creativity. Yellow sits between the two, allowing specific forms of assistance while restricting others.

The strength of the system is its clarity. Rather than treating AI as universally acceptable or universally prohibited, teachers define its role according to the learning objective.

That is also remarkably similar to how organisations will need to govern AI in the workplace. Employees will not simply receive a blanket instruction to “use AI”. They will need rules around which tools are approved, what information can be shared, which tasks can be automated and where human review remains mandatory.

Schools can begin teaching that discipline long before students enter professional life.

AI Literacy Means Knowing What Not to Delegate

The deeper value of classroom AI may come when students learn to challenge it rather than obey it.

Pillai points to an exercise used by French teacher Miriam Przybyla-Baum, in which students allow a large language model to edit their writing and then evaluate the changes. They must decide which revisions improve the work and which weaken their own voice.

That changes the role of AI entirely.

Instead of becoming an answer machine, the model becomes something students must interrogate. They compare alternatives, question whether the response is correct and defend their own decisions.

This is an important distinction because polished language can create an illusion of accuracy. A confident AI response may still contain factual errors, invented citations or reasoning that does not survive scrutiny.

Responsible AI education therefore needs to go beyond prompting. Students need to learn verification, source checking, privacy awareness and the ability to recognise when AI assistance has crossed into replacing their own thinking.

Pillai captures that tension clearly: knowing how to delegate to AI will matter, but knowing what not to delegate may matter even more.

Teachers Need AI Literacy Too

AI Literacy for Educators: Empowering Teachers to Lead with AI in Education

Students cannot carry this responsibility alone.

Teachers are increasingly using generative AI for lesson plans, assessments, presentations, rubrics and administrative work. Education-specific systems such as MagicSchool package many of these functions, while general-purpose models can perform similar tasks.

Used carefully, these tools can save educators considerable time. But they also create the same verification problem students face.

If teachers accept an AI-generated lesson plan or feedback without reviewing it, they risk modelling precisely the behaviour schools should be teaching students to avoid.

AI can produce incorrect equations, fabricate sources and simplify nuanced topics too aggressively. Automated feedback can also miss the individual abilities, context and personality of a student.

A useful principle emerges from Pillai’s argument: AI can generate, but humans still need to verify.

For schools, that means responsible AI policy cannot focus only on student misconduct. Teacher training, institutional guidance and shared standards need to develop at the same time.

Faster Is Not Always Better in Education

Much of the excitement around AI focuses on productivity. In education, productivity is not always the correct measure.

If a tool reduces a twenty-minute task to two minutes, a company may reasonably see that as an efficiency gain. In a classroom, however, those missing eighteen minutes may contain the learning itself.

Writing an essay develops the ability to organise an argument. Working through a difficult equation develops mathematical reasoning. Searching for the correct expression in another language helps build memory and fluency.

If AI removes every difficult stage, it may also remove the cognitive effort schools are trying to cultivate.

The better question is therefore not whether AI can complete an assignment, but whether using it will improve the student’s understanding.

In some situations, the answer will be yes. AI can offer additional explanations, generate practice exercises or provide alternative examples. In others, students may need to work independently first so that they develop the underlying skill before they begin delegating parts of the task.

Students Should Have a Voice in AI Governance

AI Governance at EDUCAUSE 2025: Strategy, Structure, Progress

Cheshire Academy has also experimented with a Student AI Council, according to Pillai’s article.

That idea deserves attention because young people are not simply subjects of AI policy. They are likely to become the generation most deeply shaped by these systems.

Including students in discussions about what responsible AI use should look like can turn policy into something more meaningful than compliance. They can examine where AI helps, where it creates problems and what kinds of boundaries make sense in different contexts.

That process also teaches a wider lesson about technology.

Powerful systems can have permissions, restrictions and responsibilities. Their use depends on context rather than convenience alone.

Those are governance principles students will encounter again in universities, workplaces and public life.

The Real Question Is What Kind of Thinking Schools Want to Protect

The debate around AI in education has often become trapped between two extremes: ban the technology or embrace it completely.

Neither approach reflects how AI is likely to function in the real world.

Students will encounter generative AI at university and in their careers. Pretending it will disappear does not prepare them for that future. But allowing students to delegate every difficult task to a model risks weakening the reasoning, creativity and independence education exists to develop.

The strongest approach may therefore sit between those positions.

Schools can teach students to use AI as a tool while making clear that different tasks require different levels of assistance. Some work can involve collaboration with AI. Other work should remain deliberately human.

As Pillai’s argument suggests, the most important form of AI literacy may not be technical at all. It may be the ability to recognise when a machine is useful — and when human thinking needs to remain in charge.

Source: