AI Act Education Risk Register
Product: Lumi Tutor Scope: EU education and vocational-training use. Status: Draft classification, counsel review required. Last updated: [LAST_UPDATED_DATE]
Not legal advice. This register is an internal product-control document. Final classification must be reviewed by qualified EU AI Act counsel.
Product position
Lumi Tutor is intended to be a formative tutor. It should help students practise and understand material. It must not be used to make official educational decisions unless Lumi completes a separate high-risk AI compliance program.
Classification summary
| Feature/use case | Current status | Risk position | Control |
|---|---|---|---|
| Child asks for help understanding a concept | Allowed | Lower risk when voluntary/formative and not used for official evaluation | Safety, age fit, no personal data solicitation |
| Practice questions with instant feedback | Allowed | Lower risk if feedback is not final grade or credential | Label as practice only |
| Parent dashboard summaries | Allowed | Must avoid official school decision effect | Summary-only, parent controls, no grading claim |
| Teacher dashboard progress insight | Pilot-only | Could become high-risk if used for summative evaluation | Contract and UI ban on grading/placement |
| Diagnostic or placement quiz | Family beta only | High-risk if school uses it to assign level, track, placement, or progression | Disable for school decisions |
| AI-generated grades or scores for school | Not allowed | High-risk education evaluation | Do not build until high-risk program exists |
| Admissions, school assignment, scholarship eligibility | Not allowed | High-risk | Out of scope |
| Proctoring, cheating detection, behaviour monitoring | Not allowed | High-risk and potentially biometric/behaviour risk | Out of scope |
| Special-education classification | Not allowed | High-risk and special-category/sensitive data risk | Out of scope |
| Religious/ethics content support | Pilot-only | Special-category and school-ethos sensitivity | No child belief profiling; reviewed content packs |
| University/adult learning | Not currently supported | Separate product mode required | Do not market under child Lumi |
High-risk trigger controls
The following must remain disabled or contractually prohibited for schools:
- Official grading.
- Official assessment of learning outcomes.
- Admission/access/assignment decisions.
- Level, track, placement, progression, or credential decisions.
- Proctoring or prohibited-behaviour monitoring.
- Special-education classification.
- Scholarship eligibility.
- Teacher replacement for high-stakes educational decisions.
Required product copy
Every school pilot must communicate:
- Lumi provides formative tutoring and practice support.
- Teachers and schools remain responsible for instruction, assessment, grading, placement, and safeguarding.
- Lumi outputs must not be used as the sole or decisive basis for decisions about a student.
- Lumi does not provide medical, legal, psychological, or financial advice.
If Lumi ever enters high-risk use
Before building or selling high-risk education features, Lumi needs:
- AI Act classification memo.
- Risk management system.
- Data governance and training-data documentation.
- Technical documentation.
- Logging and traceability.
- Human oversight design.
- Accuracy, robustness, cybersecurity, and bias testing.
- Transparency instructions for deployers.
- Quality-management system.
- Post-market monitoring and incident reporting.
- Vendor/deployer responsibility split.
Until then, the product rule is simple: formative tutoring only.