The rule is short: the only thing that goes out to the artificial intelligence is what you send it to analyse, and only when you ask.
What does not get sent
Your organisation's documents kept in the Library – articles of association, accounts, certificates, minutes – are never sent to the AI. They live in a private store that only your organisation sees, they cannot be reached from a public address and they are served through signed links that expire: an address copied out of the browser bar will not still be opening your articles of association a month from now. They are deleted whenever you want.
What does get sent
- Your organisation's profile, when you ask for a fit analysis. It goes as text inside the request and is not stored there permanently.
- The project you upload for [Match](/documentacion/match). That one is a file: it is uploaded to the supplier's file store and lives there ninety days at most, after which it expires on its own.
Your data is not used to train models: the supplier does not train on what is sent to it through its programming interface. Your narratives do not end up inside a model.
Even so, before uploading a narrative, take out the names and details of the people you support that the analysis does not need. With the project anonymised the result is the same.
The assistant in Help also talks to the model, but no file of yours reaches it, and neither does the content of your calls: only numbers and states – which plan you are on, how much of the profile you have filled in, how many alerts are unread.