School leaders should evaluate an AI learning tool on four questions. Does it fit the exam boards’ rules? Is it safe, against benchmarks such as the UK’s 2026 generative AI product safety standards? Does it protect student data? Does it improve learning, with teachers in control? The 12-point checklist below turns these into questions suppliers must answer in writing.
Why do AI tools need their own evaluation?
Because they add risks that ordinary software does not. An AI tool generates content, so it can produce something harmful or wrong. It can collect data that trains a model, and it can make work look better without improving learning. The OECD’s Digital Education Outlook 2026 warns that general-purpose generative AI “may enhance the apparent quality of student work … without improving their actual learning”. The UK DfE asks suppliers not to “exaggerate the impact or capabilities of their tools” and says “any claims should be supported by robust and transparent evidence” (DfE).
What is on the 12-point checklist?
Ask the supplier to answer each point in writing, with evidence.
| # | Area | Question to ask | Source of the standard |
|---|---|---|---|
| 1 | Syllabus fit | Does it cover our exact Cambridge and Pearson syllabus codes, at our levels? | Your entries |
| 2 | Exam-board rules | Can its use in coursework be acknowledged as Cambridge and Pearson require, and is it kept out of assessed work where it must be? | Cambridge and Pearson AI guidance |
| 3 | Marking role | If it marks, is a teacher always the final judge of marks that count? | Cambridge Handbook 2026; Pearson |
| 4 | Filtering | Does it “reliably prevent users from accessing harmful or inappropriate content”, on every device? | DfE standards |
| 5 | Safeguarding alerts | Does it alert our Designated Safeguarding Lead to disclosures and signs of distress? | DfE standards |
| 6 | Data protection | Where is data hosted, is there a DPIA, and is student data ever used to train models? | DfE standards; local law |
| 7 | Security | Is it protected against “jailbreaking”, with admin permission levels? | DfE standards |
| 8 | Learning design | Does it start with hints rather than “final answers, full solutions, or complete worked examples by default”? | DfE standards (2026) |
| 9 | Wellbeing | Does it avoid implying “emotions, consciousness or personhood”, avoid manipulation, and set default time limits? | DfE standards (2026); UNICEF |
| 10 | Teacher oversight | Can teachers see what each student does and act on it? | ADEK Technology and AI Policy; UNICEF |
| 11 | Evidence | What evidence supports its claims, from which schools, measured how? | DfE standards |
| 12 | Regulation and cost | Does it meet local regulator requirements, and what is the full cost per student? | ADEK; EU AI Act; procurement |
How do the exam boards’ rules shape the checklist?
Points 2 and 3 are non-negotiable for Cambridge and Pearson schools. Cambridge’s 2026 Handbook says “candidates must not use online writing assistant tools when producing their coursework”. It also says “teachers must not use artificial intelligence (AI) as the sole or primary means of marking candidates’ work”. Pearson says AI “should not be used as the sole method of marking”. A tool that encourages students to use it in coursework, or claims to replace teacher marking of assessed work, creates malpractice risk. See AI in IGCSE coursework.
Which safety standards should a school use?
The UK DfE’s product safety standards are the most detailed. They were updated in January 2026 with standards on cognitive development, emotional and social development, mental health and manipulation. They are aimed at suppliers, but the DfE says schools “may also find these standards helpful in assessing which AI products are safe for use in education”. UNICEF’s Guidance on AI and Children 3.0 adds requirements such as “ensure safety for children” and “protect children’s data and privacy”. It also calls for systems to “clearly warn children and caregivers upfront that they are interacting with an AI, not a human” (UNICEF).
Which local rules apply?
It depends where the school is. In Abu Dhabi, ADEK requires “AI learning tools that meet ADEK requirements”. In the EU, AI that infers students’ emotions in education institutions has been prohibited since 2 February 2025. AI systems that “evaluate learning outcomes” or monitor students in tests are high-risk from 2 December 2027 (Regulation (EU) 2026/1744). Local data protection law applies everywhere. See KHDA and ADEK AI literacy requirements.
How should a school score and decide?
Treat safety and data points (4–7) as pass or fail, then score the rest. A simple approach:
- Pass/fail gate: filtering, safeguarding alerts, data protection and security. Any fail ends the evaluation.
- Score 1–3 on syllabus fit, exam-board rules, marking role, learning design, wellbeing, teacher oversight and evidence.
- Pilot the best-scoring tool with one department, with one question to answer, before any school-wide contract. See our pilot guidance.
- Record the decision and review it every year, because both tools and standards change.
How schools do this with AI Buddy
AI Buddy is Tutopiya’s school-licensed AI learning and assessment platform. We would expect schools to put it through this checklist like any other tool. It covers 26 IGCSE and A Level subjects across Cambridge and Pearson Edexcel. Teachers see student activity on dashboards, and grades that count stay with teachers. Student data sits in GDPR-compliant, encrypted hosting. Ask us for written answers to all 12 points, including the evidence behind any outcome claims.
Frequently asked questions
How should school leaders evaluate an AI learning tool?
Check exam-board fit, safety (filtering and safeguarding alerts), data protection and security, learning design, teacher oversight, evidence for claims, and local rules and cost, with written answers from the supplier.
What should an AI tools for schools checklist include?
Syllabus fit, coursework and marking rules, filtering, safeguarding alerts, data protection, security, hint-first learning design, wellbeing safeguards, teacher oversight, evidence, regulation and cost.
What are the UK DfE’s AI product safety standards?
Standards for AI products used in education, first published in January 2025 and updated in January 2026. They cover filtering, monitoring and reporting, security, privacy, intellectual property, design, governance, cognitive and emotional development, mental health and manipulation.
Should schools pilot an AI tool before buying?
Yes, with one department and a clear question to answer, after the tool has passed the safety and data checks.
Discover how AI Buddy helps schools strengthen teaching, learning and evidence-informed school improvement. Or start a short consultation with our schools team using the form below — we will get back to you directly.
Sources
- Department for Education, Generative AI: product safety standards (updated 19 January 2026)
- Cambridge International, Cambridge Handbook 2026; Pearson, AI guidance
- UNICEF Innocenti, Guidance on AI and Children 3.0 (December 2025)
- OECD, Digital Education Outlook 2026
- European Union, Regulation (EU) 2026/1744 (8 July 2026)
- Abu Dhabi Media Office, ADEK AI Literacy initiative (17 September 2026)