AI Notice for Odyseya

Last Updated: October 2026

You are talking to an AI

Every reflection in Odyseya is generated by a language model, not written by a person, and no human reads your entries.

What Odyseya is not

Odyseya is a tool for self-reflection, habits and personal growth. It is not a medical device, not medical advice, diagnosis or therapy, and not a substitute for professional care. None of its companions is a therapist, doctor, or counsellor.

If you are in crisis, the app can show you local emergency and helpline numbers. That is a safety signpost, not a clinical assessment. If you or someone else is in immediate danger, call your local emergency number.

The three companions

Each one is a separate prompt to the same language model, with a different job and different access to your history.

• Jerboa, every entry, automatically. Reflects your own words back to you. He does not advise, instruct, or suggest. He is the only companion that carries memory across entries.
• Little Owl, only when you ask. Offers two or three possible directions after you have read Jerboa. She sees only that single entry and that single reflection - no history, no stored facts.
• The Camel, whenever you open My Desert, and on your Oasis board when you ask for your daily words or for help finding the feeling beneath a goal. In My Desert it reads the shape of your recent days - when you wrote, and the short summaries Jerboa made of each day (a one-sentence essence of the day, the main feeling and its strength, topics, the pattern noticed) - never the text of your entries. To see the change over time it also receives up to 15 durable facts from the facts ledger described above and its own summary of the previous month. It speaks again only once that shape has genuinely changed since it last did. On the Oasis board it receives the feeling words and meanings you wrote there, the goal beneath each feeling, the short moments saved with it, and how often you marked it as felt, or the goal you typed, together with your North Star and values.

What the model receives

Your voice is turned into text on your own phone, by the phone's speech recognition set to run on the device only. The audio is never saved as a recording and never uploaded - not to us, to Apple or Google, or to OpenAI. Only the text becomes your entry.

The text of an entry is sent over an encrypted connection to our server and on to OpenAI, which generates the reflection and returns it. Under OpenAI's API terms it is not used to train any model, and OpenAI may keep it for up to 30 days for abuse monitoring before deleting it.

Along with today's entry, Jerboa receives your North Star, the goals you set for areas of your life, the feeling words from your Oasis board, a small ledger of durable facts gathered from earlier entries (people, projects, places, ongoing situations, health, preferences), recurring themes and loops noticed so far, and may also receive your most recent earlier entries for pattern comparison. He is instructed never to quote those facts back at you. The model never receives your name, email or location.

When you set your North Star or a goal, its text is also sent to OpenAI to check that it is phrased as something within your own reach.

Reading handwriting

Cloud handwriting reading is currently switched off: handwritten pages are read on your phone only, and no photo leaves it.

What is kept

Reflections, Little Owl's directions, the Camel's readings, detected patterns and the facts ledger are saved with your account so you can come back to them. Unlike the text of your entries, which is encrypted on your phone before it is saved, these are stored on our servers in readable form, protected by Google's encryption at rest. They are deleted with your entries or your account.

What it can get wrong

A language model can misread tone, invent a connection that is not there, or reflect something back in a way that does not fit. A reflection is one reading of your words, not a verdict on you, and no decision about you - your account, access or price - is made from it. You are always the authority on your own experience.

Your consent

AI processing of journal entries runs on your explicit consent under Article 9(2)(a) GDPR, given before your first entry. You can withdraw it at any time in Settings > How the companions think > Companion reflections, which stops all future AI processing of your entries, including the Camel and reading handwriting with AI. Withdrawing does not delete reflections already generated - you can remove those with your entries.