Trust & safety
AI safety for students
The AI is built to teach, not to do the work — and to stay checkable. This page explains the guardrails: textbook grounding with page citations, hint-ladders instead of answers, a no-ghostwrite rule, teacher oversight, and an absolute no-training-on-student-data commitment.
Where do the AI’s answers come from?
From the school’s own uploaded textbooks. Answers carry page citations so a student or teacher can check every claim against the book. When a question is scoped to a chapter, the AI reads that chapter directly rather than guessing from the open web.
Can students make the AI do their work?
No — that is designed out. Socratic Practice climbs a hint ladder and never reveals the final answer; the Writing Coach comments on the student’s own draft and refuses to ghostwrite; stuck students are flagged to their teacher rather than handed a solution.
Is there human oversight?
Yes. Teachers see who is stuck and where, hold review steps such as approving song lyrics before music is generated, and can release or withhold results. The AI supports the teacher’s judgement; it does not replace it.
Is student work used to train AI?
Never. Student work, chat history, voice recordings and uploaded textbooks are not used to train any AI model — ours or a vendor’s. The AI reads material to answer in the moment; nothing a student produces becomes training data for anyone.
Questions about AI safety
Write to info@edvation.ai for the AI-safety detail your procurement team needs. See also how content and copyright are handled, and the subprocessor list.
More in the Trust Centre
Edvation