AI & safety

All of it runs on OpenAI’s API. Nothing about your notes is processed by a model we host ourselves.

• Turning speech into text, both for a recording and during a live conversation with the assistant.
• Reading the text out of a photo.
• Writing a title for a note, and generating its summary, minutes, actions, checklist and any custom tab you define.
• Answering your questions about your notes.

• A transcript can mishear a word — names, numbers, medical and legal terms, and accented or code-switched speech are the usual casualties.
• A summary can state something the recording never said, or quietly drop the one detail that mattered to you.
• An assistant answer can be confidently wrong, and can mix up which note something came from.
• Models reflect biases in what they were trained on, and they do not know what they do not know.

The recording is the evidence. The transcript is a reading of it, and a summary is a reading of the transcript — each step can lose or invent something. Where a tab and the transcript disagree, believe the transcript; where the transcript looks wrong, listen to the audio.

Do not rely on generated text for a medical, legal, financial or safety decision without checking it against the original. younote is a notebook, not an adviser and not a system of record.

We do not train models on your content, and we do not let anyone else do so. Content sent through OpenAI’s API is not used to train their models under their API data usage policies.

A transcription app makes it very easy to record a conversation, and the law about doing so differs from country to country — in many places everyone present has to agree. That judgement is yours to make before you press record, and we are not in a position to make it for you.

A bad transcript or a fabricated summary is worth reporting — it is how we learn where the app misleads people. Send the note’s title and what went wrong to support@yazool.net. Please do not send anything confidential you would not want read.