AI Safety Checklist for Teachers and Course Creators Teaching Teens Online
A practical, platform-agnostic checklist for setting safe, transparent boundaries around AI use with teenage learners. Use it before launch, in learner and parent communications, and in a short weekly review routine.
AI can support teenage learners with explanations, practice questions, feedback prompts, and accessible ways into difficult material. But a safer class does not begin with choosing a tool. It begins with a teaching plan: what AI may do, what it must not do, what information never enters it, and where an educator remains responsible.
This matters especially in online courses, where learners may work independently and families may have limited visibility into day-to-day study habits. A clear plan protects learning as well as wellbeing: students know when to think first, when to ask for help, and how to treat an AI response as a starting point rather than an authority.
OpenAI announced ChatGPT for Teens on August 18, 2026, describing additional teen protections, healthy-use features, parent controls, and learning-oriented features such as guided study support. That is a useful signal that teen use needs age-aware design. It is not, however, a substitute for an educator’s own boundaries, review processes, or safeguarding responsibilities.
Why teen AI use needs a teaching plan, not just a tool decision
Every AI-enabled activity makes an instructional choice. If a learner asks for an explanation after attempting a problem, AI may support practice. If they paste a task and submit the result unchanged, the activity may conceal whether they understand the material. The difference is not the prompt alone; it is the task design, the evidence of thinking you collect, and the adult support available when something goes wrong.
A child-centred approach also asks broader questions: Is the learner safe? Is their personal information protected? Can they understand how AI is being used? Is the activity fair and accessible? UNICEF’s guidance on AI and children identifies safety, privacy, fairness, transparency, wellbeing, inclusion, and preparation for AI as central requirements.
Use these questions to write a short AI plan before you add a chatbot, image generator, feedback assistant, or automated workflow to a teen course. Keep the plan understandable enough that a student and parent can act on it.
Pre-launch AI safety checklist for teen classes and courses
Download-ready one-page checklist: copy the table below into your course handbook or a document, add your organisation’s contacts and policies, then save or print it as a one-page handout.
| Check before launch | What to decide or do |
|---|---|
| 1. Name the learning purpose | For each AI activity, state the skill students are practising. Remove the activity if AI adds novelty but not a clear learning benefit. |
| 2. Set the age and access route | Confirm the tool’s age requirements, account process, and your school or organisation’s approved access route before asking teens to use it. |
| 3. Minimise data | Tell learners not to enter full names, contact details, passwords, grades, health information, private messages, identifiable images, or sensitive details about themselves or others. |
| 4. Define permitted help | Specify acceptable uses, such as explaining a concept, generating practice questions, suggesting a revision plan, or giving feedback on a draft. |
| 5. Define prohibited uses | Prohibit impersonation, harassment, sexualised content involving minors, sharing another person’s data, bypassing supervision, and submitting AI-generated work as wholly one’s own. |
| 6. Keep human review | Identify work that requires teacher review: final grades, high-stakes feedback, welfare concerns, conflict, accusations, and any decision that materially affects a learner. |
| 7. Build verification in | Require students to check important claims against reliable course materials or teacher-approved sources, and to flag uncertainty instead of guessing. |
| 8. Make thinking visible | Collect a plan, annotations, draft history, short reflection, oral explanation, worked steps, or source notes—not only a polished final answer. |
| 9. Create a stop-and-report route | Give learners a named adult or support channel for upsetting, unsafe, discriminatory, sexual, threatening, or otherwise concerning AI outputs. |
| 10. Tell families early | Share the purpose, boundaries, data rules, support route, and opt-out or alternative arrangements available under your local policy. |
| 11. Test the activity | Try the prompts yourself, including ambiguous and off-topic inputs. Check whether the task is accessible and whether your instructions still work when AI gives an inaccurate answer. |
| 12. Schedule review | Set a weekly time to inspect learner work, recurring problems, unexpected outputs, and whether the activity still supports the intended skill. |
Where AI can support learning—and where human oversight stays essential
A sensible default is to use AI for practice, explanation, brainstorming, rehearsal, and feedback prompts, while retaining adult judgement for decisions about learners. For example, a student might ask AI to turn vocabulary into a quiz, explain a worked example in simpler language, role-play a low-stakes discussion, or suggest questions to ask when revising a draft.
Some activities should not be delegated to AI alone. These include assessing safeguarding disclosures or emotional distress; deciding grades, placement, discipline, accommodations, or access to opportunities; resolving disputes; making claims about a student’s ability or character; and providing specialist medical, legal, or mental-health guidance. An AI tool may produce language that sounds confident without being appropriate for the individual learner or situation. Escalate concerns through your established human procedures.
UNESCO’s guidance warns that rapidly evolving generative AI tools can outpace institutional readiness and raises concerns about user data privacy and the ability of education institutions to validate tools. A practical response is to treat every output as material to review, not as an approved teaching resource by default.
Write expectations that learners and parents can understand
A long policy is not a replacement for plain language. Put a short agreement in the course welcome pack, review it live with learners, and invite questions before the first AI-enabled assignment. Avoid implying that a platform’s safety settings make every interaction safe or appropriate. Explain what the course controls and what it cannot control.
Fill-in-the-blank learner and parent AI-use agreement
Purpose: In [course/class name], AI may be used to support learning in specific activities. It does not replace my own thinking, my teacher’s judgement, or trusted adult support.
I may use AI to: [list 3–5 permitted uses].
I will not use AI to: [list prohibited uses, including submitting unacknowledged AI work and sharing personal or other people’s private information].
Before I submit work: I will [describe checking, attribution, reflection, or evidence-of-thinking requirement].
If AI gives me upsetting, unsafe, or confusing content: I will stop, save only the information needed for reporting if it is safe to do so, and contact [named teacher/designated contact] through [approved channel].
Family communication: Parents or carers can ask questions or raise concerns by contacting [contact details]. Alternative arrangements are [describe if applicable].
Agreement: Learner name/signature: [ ] Date: [ ]. Parent/carer name/signature: [ ] Date: [ ]. Educator name: [ ].
Design assignments that show student thinking
If a task can be completed by pasting a prompt and submitting the first answer, redesign it. This is not about catching students out; it is about collecting credible evidence of learning and giving students a reason to engage with the process.
- Ask for a before-and-after response: the learner’s first attempt, the AI feedback they chose to use, and an explanation of what changed.
- Require a claim check: identify one AI claim, verify it with course material or an approved source, and explain whether it held up.
- Use process checkpoints: topic choice, outline, first draft, peer discussion, revision note, and final submission.
- Add a brief oral or recorded explanation of one decision, worked step, or source choice.
- Ask for an AI-use note: what tool was used, what it contributed, what the learner rejected, and what they did independently.
These choices make responsible use visible without assuming that every AI interaction is misconduct. They also give teachers better information for feedback and help students practise judgement—the skill that remains valuable even when a tool can generate fluent text.
A weekly review routine for AI-enabled learning
- Review samples: inspect a small, representative set of prompts, outputs, and learner reflections.
- Look for patterns: identify repeated inaccuracies, biased examples, privacy mistakes, over-reliance, or confusing instructions.
- Check participation: ask whether the activity is helping all learners participate, including those who lack access, confidence, or a suitable account.
- Update one thing: revise a prompt, add an example, narrow an allowed use, or remove an activity that is not serving the learning objective.
- Communicate changes: tell learners and families what changed and why, in plain language.
SubSchool can help you turn these recurring teaching decisions into reusable checklists, assignment instructions, parent communications, and review prompts—while you retain authorship and make the final educational decision. Explore SubSchool to build a repeatable AI-use workflow around your own course standards.
Sources used
Sources and methodology
Prepared from the supplied editorial brief and three primary or authoritative sources: OpenAI’s August 18, 2026 announcement on ChatGPT for Teens, UNESCO’s guidance on generative AI in education and research, and UNICEF Innocenti’s child-centred AI guidance. The article intentionally gives platform-agnostic instructional practices rather than legal, safeguarding, or product-compliance advice. Recommendations are framed as educator-controlled procedures; local institutional requirements and tool terms must be checked separately.
Use the relevant SubSchool workflow while keeping the result editable and source-grounded.
