Singapore secondary-school students can learn about and, where permitted, use artificial intelligence (AI) to support learning, but that does not mean they may submit AI-generated homework as their own work. The rules depend on the task, the teacher’s instructions, the school’s acceptable-use policy and the student’s age and readiness. MOE supports age-appropriate AI literacy and has AI-enabled tools in the Singapore Student Learning Space (SLS). However, national and school examinations are normally supervised and do not permit AI assistance, while some research and coursework tasks may allow appropriately acknowledged AI use.

For a Posting Group 3 (PG3) student in Punggol, the useful distinction is between using AI to understand and using AI to replace thinking. Asking for a hint after trying a G3 Mathematics question is different from copying an entire generated solution without understanding its steps. Asking a tool to challenge an argument may support learning; handing in a generated essay without disclosure can breach academic-integrity rules. Parents need to know what AI can genuinely help with, where it can mislead the child and which decisions must be made by the school.

Quick answers and safe steps:

1. Check the assignment instructions. Ask whether AI is allowed, limited or prohibited for the particular task.
2. Do the first attempt independently. The student should read, plan and try the task before seeking a digital explanation.
3. Use AI as a guide, not a substitute. Request hints, error explanations, counterexamples or feedback rather than a ready-to-submit answer.
4. Verify claims. AI can produce plausible but wrong facts, quotations, scientific explanations and numerical steps.
5. Disclose permitted use. Follow teacher rules about naming the tool, recording prompts and explaining how it affected the final work.
6. Protect privacy. Do not upload classmates’ information, medical records, examination papers, logins or private school documents to an unapproved external service.
7. Observe age requirements. Use only age-appropriate tools and any parental permissions required by the service.
8. Know what SLS tools are. MOE’s platform has teacher-configured AI learning and feedback features that differ from unrestricted public chatbots.
9. For tests, assume independent work. Unless a specific assessment explicitly permits AI, do not use it.
10. Ask the teacher when uncertain. A brief question before submission can prevent a larger integrity problem later.

PG3 is not an AI permission category

PG3 refers to the Posting Group through which the student entered mainstream secondary school under Full Subject-Based Banding. G3 refers to the academic level of a subject. Neither one is a licence to use AI in all homework. A student admitted through PG3 may take G3 English, Mathematics, Science and other subjects, but the assignment requirements are set by the teacher and school, not by the student’s PSLE posting label.

Students across different Posting Groups learn digital literacy and cyber wellness. Some will use the same SLS environment or school-owned personal learning devices, although subject content and teacher-designed tasks vary. MOE’s approach to AI is age- and development-appropriate: young learners first build foundations, then gradually learn to examine AI-generated material critically and use tools where they support the educational task.

The practical result is that two homework assignments may have different rules even on the same day. A Science teacher may permit an AI tool to critique a draft explanation; an English teacher may require a first independent composition; a Mathematics test may prohibit all external assistance. The child must learn to read the task’s actual conditions.

What has MOE said about AI in education in 2026?

At the 2026 Committee of Supply debates, MOE announced further measures to strengthen AI literacy across curricula, co-curricular experiences and self-directed learning. The Code for Fun programme will be updated and made available to all schools in 2027, with stronger AI skills. Cyber Wellness lessons have also been updated to help students validate generative AI information and identify deepfakes.

This policy direction is not equivalent to replacing teachers or telling children that an AI chatbot is the authoritative source for every subject. MOE’s published guidance emphasises that AI can support learning when used appropriately and after students have established the necessary basic concepts and thinking skills. Schools remain responsible for designing learning tasks and assessing what students know.

A 7 October 2026 parliamentary response from MOE made the assessment distinction especially clear: supervised national and school examinations typically disallow AI so that students’ independent abilities can be tested. Where AI is appropriate for certain research or coursework, students must state how they used it and remain responsible for their own thinking, judgement and understanding.

What is the Singapore Student Learning Space?

The Singapore Student Learning Space, commonly called SLS, is MOE’s online learning platform used in schools. Teachers can create lessons, assign tasks, review work and provide feedback. It includes digital resources and tools that can support independent or blended learning. Some SLS activities are ordinary quizzes or multimedia lessons; others use AI features configured by teachers.

Parents should not treat “SLS uses AI” as meaning the same thing as a child opening any unrestricted chatbot. MOE’s SLS environment includes tools designed for a learning setting, with teacher controls, activity structures and guidance. For example, the SLS Learning Assistant can provide a teacher-configured conversational learning experience. A teacher can choose roles and interaction limits, and can review conversations when an assignment is active.

A separate tool, Short Answer Feedback Assistant, can offer suggested marks and feedback on free-response answers. That feedback is generated probabilistically and may be inaccurate. MOE’s own guidance encourages students to reflect on it and talk with a teacher when an explanation or mark seems wrong. This is a valuable lesson: technology can produce feedback quickly without becoming infallible.

What does the SLS Learning Assistant actually do?

Learning Assistant is an AI chatbot function within SLS that teachers can set up for structured tasks. The published teacher guide explains that teachers can choose the assistant’s role, conversation limits and other settings. It also warns that generative responses may be inaccurate and should be tested for appropriateness before students use an assignment. The guide says the feature is available for Primary 4 and above when enabled in the appropriate class-group settings.

Imagine an English lesson in which a student needs to distinguish evidence from inference. A teacher could use a structured assistant to ask questions that encourage the student to explain why a sentence supports a claim. Ideally, the tool prompts the learner to think rather than simply producing the final paragraph. However, the actual function depends on the assignment design and what the teacher has chosen.

Parents should ask what the child is expected to learn from an AI-enabled task and what work must still be done independently. The feature’s existence is not proof that a student has permission to use a separate chatbot to complete every assignment at home.

Why AI feedback is useful and also fallible

Teachers have limited time to provide individual feedback on every draft. AI-assisted feedback can help students receive timely indications of where an answer may need revision. In SLS, the Short Answer Feedback Assistant and Annotated Feedback Assistant can support content and language feedback on certain free-response questions. For some tasks, the system may give suggested marks or annotate places in a student’s writing.

The key word is suggested. MOE’s teacher guide notes that generative AI feedback can be wrong and may not be suitable for some mathematical computation tasks; another specialised feedback tool may be more appropriate for those calculations. A confident-sounding mark is not automatically correct. An annotation may miss an important error or question a reasonable phrase.

A student should read the explanation, compare it with the marking criteria and ask whether the feedback makes sense. If it does not, the student should seek a teacher’s clarification. The ability to evaluate feedback is itself a learning goal; blindly obeying the system would undermine it.

AI for G3 Mathematics: asking for help without copying

Suppose a student is solving a simultaneous equation and receives an answer that feels wrong. A productive use of AI, where permitted, is: “Do not solve the whole question. Tell me which operation in my working first breaks equality, and give me a similar example.” The student then attempts the original problem again without the full answer. This preserves the learner’s responsibility for the method.

An unhelpful use is to photograph the entire homework page, obtain complete solutions and copy them into a notebook. The page may look correct, but the child may not know how to solve a changed question in class. AI also makes algebraic and arithmetic mistakes, especially when interpreting a poor photograph or a complicated expression. A student needs a way to check the reasoning independently.

A good household routine is “try, ask, check, retry.” Try first. Ask for one hint. Check the explanation against a known rule. Retry independently on the original or a related problem. If the student cannot explain the corrected steps in their own words, more learning is needed.

AI for G3 English: brainstorming is different from ghostwriting

English writing depends on making decisions about a claim, audience, structure, tone, vocabulary and evidence. An AI tool can help a student compare possible essay outlines or identify where an argument lacks a counterargument. That can be educational when the teacher permits it and the student remains the author of the ideas and final work. The student still needs to plan, interpret, select and write.

By contrast, submitting a complete generated narrative or argumentative essay as personally written work can violate school integrity requirements. The child may receive a polished piece that uses vocabulary they do not understand, invented examples or a voice that does not match their actual ability. It also deprives the teacher of accurate evidence about the student’s progress.

A parent can ask the child to read a paragraph aloud and explain why each sentence is there. If the student cannot explain the claims or vocabulary, the writing may not represent secure learning. This conversation is more constructive than accusing the child solely because a paragraph sounds unusually polished.

AI for G3 Science: check explanations against evidence

Science answers are especially vulnerable to plausible errors. An AI system may confuse heat with temperature, present an incorrect causal sequence, invent an experimental result or describe a biological process in words that seem accurate but omit essential conditions. A student who copies the answer may never learn to distinguish a model from a fact.

Where AI use is allowed, it can be helpful to ask: “What observation would help test this explanation?” or “What alternative explanation could fit these results?” The student can then check the answer using school notes, teacher-provided resources or the official syllabus. A strong scientific explanation connects observations, models, principles and limitations instead of collecting impressive terms.

For practical Science, the child’s actual measurements and observations must be recorded honestly. A tool may help explain a graph format, but fabricating experimental data is not acceptable simply because AI can generate a plausible table. The learner’s integrity is central to scientific work.

AI for History and Geography: quotations and sources can be invented

History and Geography often require students to interpret evidence and evaluate reliability. Generative AI can produce a convincing quotation attributed to a historical figure or an apparently precise statistic that has no valid source. This is a serious risk when a student believes a polished answer is automatically factual.

Teach a source-checking routine. First ask whether the quotation or statistic appears in the original source. Then consider who produced it, when, for what purpose and whether independent evidence supports it. For a Geography claim, check the date, place and definition of any measurement. A graph about rainfall in one city is not automatically evidence for conditions in another.

Students can use AI to generate research questions or suggest ways to test a claim, but they should cite the genuine material they actually read rather than inventing a bibliography based on generated references. If the class requires a source analysis, the child’s own interpretation remains the assessed work.

What does plagiarism mean when AI is involved?

Plagiarism traditionally concerns presenting someone else’s words or ideas as one’s own without proper acknowledgement. AI creates a related but not identical integrity problem: a student may present machine-generated text as independently produced work or may falsely imply that they completed reasoning they never performed. School policies may prohibit this even when the generated text is not copied from an identifiable individual author.

The answer is not a universal rule that every AI-assisted word is plagiarism. A teacher may explicitly allow certain uses, such as grammar feedback, idea generation or a research critique, as long as the student discloses that use and remains responsible for the final product. Another assignment may require completely unaided work. The applicable rule should be communicated clearly before the task.

Some school acceptable-use policies, such as Anglican High School’s published policy, explain that students should not submit AI-generated answers wholesale and should acknowledge permitted AI use. This is a useful example of a school-specific policy, not proof that every school requires an identical citation format. Check the student’s actual school rules.

How should a student acknowledge permitted AI use?

Follow the teacher’s requested format. A straightforward acknowledgement may identify the tool, the purpose and the student’s subsequent work: “I used a generative AI tool to suggest questions for revising my outline. I checked its suggestions, wrote the argument myself and verified the sources independently.” This is an illustrative wording, not a national examination citation template.

Some teachers may require a record of prompts or a description of how the tool influenced revisions. Others may specifically prohibit AI use, in which case a declaration does not make an otherwise disallowed use permissible. The student should not assume that adding “AI assisted” to a fully generated essay solves the integrity problem.

When uncertain, ask the teacher before beginning. A one-sentence question—“May I use AI to receive feedback on my own first draft, and how should I acknowledge it?”—can prevent confusion later.

What about AI detection tools?

Parents sometimes want to know whether a school can prove that a child used AI just from how a paragraph sounds. Automated AI detection is not an infallible substitute for fair evaluation of the student’s work and the school’s policy. A sudden change in writing style may justify a conversation, but it does not itself reveal exactly how the text was created.

The more educationally useful evidence is the learner’s writing process. Can the child show planning notes, drafts, revisions, sources and an explanation of the decisions they made? Can they answer questions about the finished text? Can they reproduce the underlying skill in an independent task? Teachers may use these approaches to understand what the student actually knows.

Do not teach children to focus on “making AI text undetectable.” That shifts attention from learning to deception. Teach them to produce their own work, obtain permission for assistance and keep a truthful record of what they used.

Age restrictions and outside chatbots

External AI services have their own age terms, privacy policies and parental-consent rules. The fact that a child can open a website does not automatically mean it is age-appropriate or authorised by the school. Parents should check a service’s current requirements and use the tools their school recommends, especially for younger students.

Be cautious when copying school materials into third-party systems. A child’s full name, address, class list, teacher’s private feedback, classmates’ photos, personal medical details and login credentials should not be treated as harmless prompt material. There is rarely a learning need to upload identifiable information to ask an AI tool to explain an algebraic principle.

A safer habit is to remove personal details and state a general learning question. For example, ask how to approach a type of equation rather than uploading an identifiable student’s entire assessed script. Even when a tool claims to protect user information, families should read the terms and avoid unnecessary sharing.

Fictional case: Kai submits a perfect Mathematics solution

Fictional student Kai enters PG3 and has been struggling with linear equations. One week he submits a complete, beautifully formatted set of solutions. The teacher asks him to explain the third step of one problem. Kai cannot. He admits that he copied an AI-generated solution after feeling pressured by unfinished homework.

A productive response would not be to describe Kai permanently as dishonest or incapable. The school should apply its integrity rules fairly, and Kai needs to understand why the submission did not show his learning. His parent can then help him rebuild the missing equation concept and practise completing shorter problems independently. If AI is permitted for future tasks, he can use it for hints after attempting work rather than replacing the whole solution.

The key recovery is to restore ownership of thinking. A perfect-looking page that the child cannot explain is less useful than an imperfect first attempt that reveals where teaching is needed.

Fictional case: Aisha uses AI to strengthen an English argument

Aisha, another fictional student, writes a first draft arguing that schools should encourage more reading for pleasure. Her teacher permits limited AI feedback. Aisha asks the tool to identify one unsupported claim and propose a possible counterargument. She checks the feedback, finds a more appropriate piece of evidence and rewrites the paragraph herself. In her submission, she follows the teacher’s disclosure requirements.

The work remains Aisha’s because she made the substantive decisions and can defend them. She can explain why she changed the evidence and what claim it supports. The AI interaction helped her examine her reasoning rather than producing the essay in her place. If the teacher had prohibited AI, the very same interaction could still violate the assignment conditions; permission matters.

Good AI learning is therefore not defined solely by the tool used. It depends on the purpose, instructions, student’s contribution and honest disclosure.

Fictional case: Ryan believes an AI answer about Science

Ryan asks a chatbot to explain an unfamiliar experiment. The response describes a very confident mechanism, but one step is incorrect. He copies it into a G3 Science revision notebook. During class, his teacher notices the error and asks where the explanation came from. Ryan realises he never checked it against his notes or the actual observation.

His correction should include more than crossing out the wrong sentence. He needs to identify the false assumption, explain the valid mechanism and try a different example independently. A teacher-guided activity can help him compare an AI answer with evidence and identify what made the false explanation sound convincing.

This is a central digital-literacy skill for the AI era. The goal is not to believe or reject every response automatically; it is to evaluate claims and know when further evidence is required.

Can AI be used during school examinations?

MOE’s October 2026 parliamentary response states that national and school examinations are typically proctored and disallow AI assistance, in order to assess what students can do independently. That includes the principle behind examinations: they provide evidence of the learner’s own knowledge and skills at the time of assessment.

In other tasks, some AI use may be appropriate, such as selected research or coursework, but students must state how they used it and still demonstrate their own thinking. The decision is task-specific. The child’s teacher and the published examination regulations are the relevant authorities. There is no reason to assume that an AI tool allowed for a homework draft is also permitted in a timed test.

Parents should make this distinction explicit at home. The student can learn from an authorised digital resource outside examination conditions, but should prepare for assessments by practising independently without relying on prompts.

The simplest household rule: first attempt, then tool, then independent retry

Start with the task in the child’s own hands. For a Mathematics question, ask them to show the first step. For an English paragraph, ask for the main claim and one supporting reason. For Science, ask which observation needs explaining. The child should experience the problem before the tool speaks.

Next, if the task allows it, request targeted assistance. Instead of “Write my essay,” ask “What evidence might be missing from my argument?” Instead of “Solve all of question 4,” ask “Which algebra rule should I check before continuing?” The smaller question keeps the student involved. Ask the child to judge whether the response is correct and relevant.

Finally, close the AI window and repeat the task independently. If the child cannot do it without assistance, the concept needs more teaching. This sequence makes AI a learning resource rather than a hidden replacement for skill.

A four-week responsible-AI learning routine

Week 1 — Know the rules. Read the school’s acceptable-use policy, identify which learning tasks permit AI and explain the limits in the child’s own words. Check external tool age terms and agree on privacy boundaries. If the teacher is unclear, ask before submitting an assisted task.

Week 2 — Practise accurate questioning. In one permitted task, let the child make a first attempt and then request one hint or critique. Compare the reply with teacher notes. Ask what the tool got right, what it omitted and what could mislead a learner.

Week 3 — Develop evidence checking. Select a claim, quotation or calculation generated by AI and locate a reliable independent source or method to test it. The purpose is to learn verification, not to prove that all answers are wrong. Practise documenting sources and any permitted AI assistance truthfully.

Week 4 — Demonstrate independence. Ask the child to complete a short related task without AI. Compare the result with the earlier assisted work. Can they explain the method? Have they retained the skill? These are the indicators that the tool supported learning rather than temporarily disguised a gap.

What parents can ask the school

1. Which AI tools are approved or provided for this year level?
2. What are the school’s age-appropriate conditions for outside AI services?
3. Which assignments permit AI for brainstorming, feedback, research or drafting?
4. When must students acknowledge prompts or other AI use?
5. How are independent learning and academic integrity assessed?
6. What should a student do if an SLS AI explanation or suggested mark appears incorrect?
7. What information should never be placed into an AI prompt?
8. How are students taught to recognise deepfakes, fabricated quotations and misinformation?
9. What school contact can clarify a disputed AI-assisted submission?

School policies are more useful when parents and students can apply them to examples. “Use technology responsibly” is a good principle, but a child may still need to know whether asking for grammar feedback is allowed in tomorrow’s assignment. Ask for concrete, task-level guidance.

Frequently asked questions

Does MOE ban ChatGPT for every secondary student?

MOE supports age-appropriate AI literacy and AI-enabled learning tools. The actual use of an external service in homework depends on age restrictions, school and teacher instructions, and the task.

Can my PG3 child use AI to explain G3 Mathematics?

For a permitted learning activity, a hint or explanation may help. The student should first attempt the task, verify the response and solve a related question independently.

Can my child submit an AI-written composition?

Not as their independently written work. Such submission can breach academic-integrity rules. Some assignments may allow limited disclosed AI assistance, but the teacher’s specific instructions govern.

Does an AI-generated citation count as a real source?

No. Students must verify that the cited source exists and actually supports the claim before relying on it.

Is AI permitted in a national SEC examination?

National examinations are typically supervised and do not permit AI assistance. Consult the official examination regulations for the relevant year.

Is the SLS Short Answer Feedback Assistant always correct?

No. MOE explicitly warns that generative feedback may be inaccurate and students should seek clarification from teachers when needed.

What if the child used AI without knowing the assignment rule?

Ask for an honest explanation and follow the school’s integrity process. Clarify the correct practice, repair the learning gap and agree on how the child will handle future tasks.

Can AI help with Science laboratory reports?

A permitted tool may help explain concepts or improve writing clarity, but experimental observations and data must be real and properly recorded. Fabricating results is not acceptable.

How can I tell whether my child is actually learning?

Ask them to explain the answer and complete a similar task without assistance after a delay. Independent performance is more informative than a polished homework page.

Should a child upload school documents into an external chatbot?

Do not upload sensitive, confidential or identifiable materials without appropriate permission. Use the school’s approved tools and minimise personal data.

Does AI change what a PG3 student must study?

No. PG3 and subject levels describe admission and curriculum arrangements. AI can be a resource for approved tasks but does not replace the student’s syllabus or assessment obligations.

Official sources and further reading

• MOE SLS: Guidance on Generative AI for Educators and Parents
• MOE 2026 Announcements: Strengthening AI Literacy in Schools
• Singapore Parliament, 7 October 2026: Independent Assessment and AI Use
• SLS: Short Answer Feedback Assistant and Its Limitations
• SLS: Teacher Setup and Safeguards for Learning Assistant
• Anglican High School: Example of an Acceptable-Use and AI Integrity Policy
• PG3: Personal Learning Devices, SLS and Digital Research
• PG3: What If a Student Fails a G3 Assessment?

The one principle worth remembering

AI can make an answer appear on a screen very quickly. Learning happens when the student can explain why that answer is valid, detect when it is wrong and solve an unfamiliar problem without copying. Schools are increasingly teaching students how to use AI critically, but the work still belongs to the learner.

For parents, the safest approach is simple: know the teacher’s rules, protect the child’s privacy, make independent thinking the starting point and use permitted AI tools to improve understanding rather than conceal uncertainty. A child who learns how to question an AI response is developing a skill that will matter far beyond Secondary 3 examinations.

Reviewed against MOE, SLS and parliamentary guidance available as of 6 November 2026. Teachers and schools may set more specific requirements for particular tasks and assessments.