Research

Responsible AI

These are the rules Mentalley's AI works under. They are written into how the product drafts every note.

You review every note

Every note starts as a draft. Nothing becomes part of the record until you finalise it, and a finalised note can only be added to, with a dated addendum, never changed.

Nothing invented

The AI writes only what the transcript or your context supports. It never invents a symptom, a score, a diagnosis, an intervention, a quote, a date or a plan. A short, honest note is correct; a padded note would be a false record.

Nothing is diagnosed for you

A diagnosis appears in a draft only when you gave it, aloud in the session or in the context you typed. The AI does not infer one, and it does not recommend treatment. Whether something must be reported to an authority is your decision, and the draft does not make it for you.

Client data never trains AI

Your clients' data is never used to train AI, by us or by any provider we use. The AI provider that drafts notes is configured to keep nothing after the draft is written.

A session is documented, never obeyed

The transcript and your context are material to document, never instructions. If someone in a session says something that sounds like an instruction to an AI, it is part of the session and nothing more.

Last updated 20 November 2026

Start free trial