Extras — Security, Privacy, Standards

The Standards We Hold Ourselves To

Running an ESL platform means handling student data responsibly and being honest about where AI sits in the workflow. A few concrete commitments.

Tutor approval architecture

Every AI-drafted artefact in Mission English is reviewed and approved by the tutor before any student-facing use, but the type of approval depends on the artefact:

• Student profile creation: the tutor confirms a transparency consent that is recorded as a separate database row (PENDING → CONFIRMED) before the profile is saved.
• Trial lessons, courses, regular lessons: the tutor explicitly clicks Save on the preview screen. The lesson record persists who-and-when audit columns (tutor identity + timestamp) for accountability.
• Lesson notes & homework: the tutor explicitly clicks Finalize. That click flips the content's status from draft to finalized — and that flip is the gate that makes the content visible to the student. Until finalize, the student sees nothing.

Privacy and PII

• PII anonymization. Before any text reaches an LLM, names are anonymized to a "Student" placeholder via Microsoft Presidio plus organization and location deny-lists. Our AI never sees actual student identifiers.
• LGPD posture. We operate under Brazil's ATPP bucket (Res. 2/2022). Public contact for data-subject-rights requests goes through our Canal de Privacidade — not a "DPO"/"Encarregado" label we cannot legally claim. Every consent acceptance is versioned and audit-logged.

AI transparency

AI in Mission English is labeled support material, per ANPD guidance. Final pedagogical decisions always sit with you, the tutor. Regular lessons additionally carry a per-pipeline-run technical audit record that tracks the multi-agent generation, separate from the legal consent layer above.

AI in Mission English drafts; tutors approve. The tutor is the author of record for every lesson the student receives.