Safety & student data
How The Infinity Lab handles safety and student data.
This page is for school leaders, safeguarding leads and IT teams. It describes how the platform works today: who can use it, how parental consent works, how the AI coach is set up, what happens when a student may be at risk, and what data we store.
Where something isn't built yet, or is only planned, we say so.
Last updated
Who can use The Infinity Lab
- Schools create student accounts. Students can't sign themselves up.
- Teachers are invited by their school's administrator.
- "Sign in with Google" works only for email addresses already on the school's list. A school can also limit it to its own email domain.
- Account setup links expire after 72 hours.
Parental consent
- Students at or below the consent age your school sets need a parent's or guardian's consent before they can set up an account. We'll confirm that age with your school.
- The school gives us the parent's or guardian's email address. They receive an email asking for their consent, and choose "Give consent" or "Don't give consent".
- Reminders are sent after 7 and 21 days. A request that isn't answered expires after 30 days, and the student can't use the platform.
- A school that has already collected consent itself can record that instead. The record shows which member of staff confirmed it, and when.
- If a parent or guardian declines, the student can't use the platform.
- To withdraw consent, a parent or guardian contacts the school.
- Not yet Parents and guardians can't withdraw consent themselves.
Matched to each student's age
- The school records each student's age band. We don't ask for a date of birth.
- Students see only the courses and challenges set for their age band.
- The AI coach is told the student's age band, and its instructions are written for learners aged 9 to 16.
How the AI coach behaves
- The coach's main instruction is to teach rather than do the work: when a student asks for an answer, it is told to guide them instead. This is an instruction to the AI model, not a technical block, so it can't be guaranteed.
- Every student message and every AI reply is checked by a content filter (Amazon Bedrock Guardrails). It covers sexual content, hate, violence, insults, misconduct, attempts to manipulate the AI, personal details and self-harm. Flagged messages are blocked and not stored.
- If the filter isn't available, the message is blocked rather than let through.
- Instructions written by course authors can't override these platform rules.
- The AI isn't given the student's name.
Safeguarding
- If the content filter detects signs of self-harm or suicide risk in a student's conversation with the AI coach, the session is paused straight away.
- The school's designated safeguarding contact is emailed; if none is set, the school's administrators are emailed instead. The student's teachers and the school's administrators also get an alert in the app. Alerts don't include what the student wrote.
- Staff see open cases in a safeguarding queue, with a target response time of one business hour, and can lift the pause.
- Not yet The student sees a message that the session is on hold, but no helpline information.
- Not covered Journal notes, reflections and uploads aren't checked for signs of self-harm.
What teachers can see
- Teachers see the progress and work of the students assigned to them. School administrators see all students in their school.
- Teachers see the AI conversations a student has chosen to save. Students are told this, and can delete a saved conversation.
- Students' work isn't shown to other students.
- Badges and scores aren't ranked or compared between students.
What students can publish
- Students can create activities and submit them to their school's library. Each submission is screened automatically, then approved or rejected by a teacher.
What we store
- Student details: name, school email address, school and age band.
- Consent: the parent's or guardian's email address.
- Learning: work, reflections, quiz answers, uploaded files, notes, certificates and badges.
- AI coach: conversations, kept so the coach can pick up a session where it left off.
- Activity: how many times a student has signed in, and when.
- We don't collect date of birth, gender, home address or profile photos. Voice input is turned into text, and the audio isn't kept.
Where data is processed
- Today the platform runs on Base44. We're confirming its hosting region with Base44.
- Planned Moving the platform to Supabase in Singapore, before launch on 13 November 2026.
- AI processing runs on AWS in the EU, using Anthropic's Claude through Amazon Bedrock. Voice features run on AWS in London.
- Our service providers are Base44 (which runs the platform today), AWS (AI and voice), Resend (email), DocRaptor (certificate PDFs; it receives the student's name) and Google (sign-in). After the move, Supabase and Vercel take Base44's place.
Keeping and deleting data
- Students can download their work as Word files. Schools can download one student's work or a whole class's. Download links expire after 7 days.
- Students can delete their own saved AI conversations and notes.
- Not yet Automatic deletion after a set period. Retention periods still need approving, and the deletion process finishing, before we can promise it.
- Not yet Full deletion of an account on request. It needs building before we can promise it.
Records and accountability
- We log who did what, and when, for consent, role changes, staff viewing student notes, downloads and safeguarding alerts.
- The log can only be added to, not edited, and it is checked every hour.
Questions about safety or student data? Email admin@theinfinitylab.ai.