Data protection

Sub-processors

When you use ReviseWell through a school, your school is the data controller and ReviseWell acts as a data processor. To deliver the service we use the sub-processors below. Schools are notified of changes before a new sub-processor starts processing personal data.

Answers go to AI marking without a name

Written answers are sent for AI marking with no student name attached, and providers are on no-training terms.

Analytics carry no answer text

Product analytics record usage events only — never answer content — and can be declined via cookie consent.

Sub-processor	Purpose	Data processed	Location	Notes
MongoDB Atlas (MongoDB, Inc.)	Primary database	Accounts, student answers, scores, feedback, progress	Being confirmed (EU move under review)	Encrypted at rest and in transit
AI marking backend	AI marking of written exam answers	Answer text + mark scheme — no student name attached	Being confirmed	The largest data flow. Pseudonymised; no-training terms
OpenRouter, Inc.	Routes prompts to AI models (teacher diagnosis, model answers, misconception clustering)	Answer text + mark scheme — no student name attached	United States	Routes to Anthropic (Claude) and OpenAI; no-training providers
OpenAI, L.L.C.	Speaking-assessment audio transcription and marking	Student audio + transcript — no student name attached	United States	API tier — not used to train models
PostHog, Inc.	Product analytics	Usage events (no answer text), tied to a user id	United States (EU region under review)	Declinable via cookie consent
Google Cloud Storage (Google LLC)	Hosting of question images	Question/exam images — no student personal data	United States / global	—
Stripe, Inc.	Payment processing	Billing contact + payment data	United States / global	Independent controller for payment data
Application host (VPS)	App server + roster database + marking logs	The above, at rest	Being confirmed	TLS / nginx
Email provider	Transactional email (logins, notifications)	Recipient email + message	Being confirmed	—

MongoDB Atlas (MongoDB, Inc.)

Primary database

Data: Accounts, student answers, scores, feedback, progress

Location: Being confirmed (EU move under review)

Notes: Encrypted at rest and in transit

AI marking backend

AI marking of written exam answers

Data: Answer text + mark scheme — no student name attached

Location: Being confirmed

Notes: The largest data flow. Pseudonymised; no-training terms

OpenRouter, Inc.

Routes prompts to AI models (teacher diagnosis, model answers, misconception clustering)

Data: Answer text + mark scheme — no student name attached

Location: United States

Notes: Routes to Anthropic (Claude) and OpenAI; no-training providers

OpenAI, L.L.C.

Speaking-assessment audio transcription and marking

Data: Student audio + transcript — no student name attached

Location: United States

Notes: API tier — not used to train models

PostHog, Inc.

Product analytics

Data: Usage events (no answer text), tied to a user id

Location: United States (EU region under review)

Notes: Declinable via cookie consent

Google Cloud Storage (Google LLC)

Hosting of question images

Data: Question/exam images — no student personal data

Location: United States / global

Stripe, Inc.

Payment processing

Data: Billing contact + payment data

Location: United States / global

Notes: Independent controller for payment data

Application host (VPS)

App server + roster database + marking logs

Data: The above, at rest

Location: Being confirmed

Notes: TLS / nginx

Email provider

Transactional email (logins, notifications)

Data: Recipient email + message

Location: Being confirmed

Entries marked in amber are being finalised — we publish the exact legal entity and hosting region rather than an estimate. If run pseudonymised (initials or usernames), none of the above hold directly identifying personal data. For data-protection questions or a copy of our Data Processing Agreement, email privacy@revisewell.net.