Privacy at Minerva Learn

Minerva Learn keeps what you practise so that tomorrow can start where today stopped. This page says exactly what that means: what we keep, what we never keep, and how to take it all back.

What we keep

When you use Minerva Learn from ChatGPT or from the practice player, we keep only what you ask to be kept:

Nothing is saved without your yes, or without you pressing a save button yourself.

How your program is made

Your program and each session's practice are composed on our server by an AI model: Anthropic's Claude, reached through OpenRouter. To do that, the model is sent what the program needs and nothing more — the skills you named, what you said about your time, place and things, what you asked to change, the program so far, and the lines, answers, picture captions and practice results you chose to keep. It is never sent your email, your name, your pictures, clips or recordings, or anything that identifies you. We ask for no data retention: the request is used to answer and is not kept or used for training by the model's providers. What comes back is checked by our server and kept as your program.

The pictures in your lessons — the one that opens a practice, and the ones that show what a good result looks like — are made for that lesson by an image model: Google's Gemini, reached through OpenRouter. It is sent only a description of the picture's subject, written by our own model from the lesson itself (a bowl, a jug, a pair of hands), and the house style it is drawn in. It is never sent your pictures, clips or recordings, your name, your email or your own words. Each picture is kept with your lesson and can be deleted with it.

Pictures, clips and recordings

You can share a photo, a short video, an audio recording or a drawing to learn from: the result you want to make, a page to learn from, or how yours came out. We are asked before anything is kept.

The microphone

Some practices offer a live pitch reading: you tap, the player listens for up to twenty seconds, and the reading is worked out inside your browser. No sound from the microphone is saved, uploaded or sent to us. If you'd rather not, record yourself on your phone and share the file, or simply describe how it went.

There is no camera in the practice player. It never turns one on.

What we never do

ChatGPT

When you use Minerva Learn inside ChatGPT, your conversation itself belongs to ChatGPT and is covered by OpenAI's own privacy policy. Minerva Learn receives only what its tools are given — a short note of what each request is for, the details of the practice, and the files you chose to share — and keeps only what is described above. Minerva Learn sends no reminders itself; the daily task you can set up in ChatGPT is ChatGPT's own scheduled task, and it lives with ChatGPT.

Where it lives

Your learning is stored on Minerva Learn's own server, separate from every other Minerva product: its own accounts, its own data, its own storage. The server runs on Heroku (Salesforce) in the European Union; your records and the pictures made for your lessons are kept in a Postgres database run by Neon in London, United Kingdom. Kept previews of your pictures and rendered practice materials stay on the server's own disk. Sign-in emails are delivered by Resend; they carry only the code. Programs and practices are composed through OpenRouter by Anthropic's model, and lesson pictures are made through OpenRouter by Google's image model, as described above.

How long, and taking it back

Everything you keep stays until you delete it. You can:

Contact

Minerva Learn is operated by Minerva Nova Limited, a company registered in England and Wales (company no. 17082360), 3rd Floor, 86-90 Paul Street, London EC2A 4NE. Questions, or a request about your data: privacy@minervanova.ai.

Version 2.1 · Effective 13 November 2026 · Terms of Service