AI and your data

What does moj use AI for?

For five things in your family space. All five are about turning a sentence in your own words into something structured:

• understanding a message and sorting it
• reading a receipt
• turning a voice message into text
• answering a question about your own entries
• suggesting steps for a goal or an appointment when you ask for them

There is also the box on our help page where you can ask the AI a question. It answers from the help pages and sees nothing in your family space.

The provider we use is OpenAI.

moj decides everything else by fixed rules, and we built it that way on purpose. Whether two appointments overlap. When a task comes round again. What “yes” to a follow-up question means and what “undo” takes back. Who may see what. In all these places the app works it out instead of guessing. You can look a rule up, and it gives the same result tomorrow as today.

What goes out with a message?

The message itself, along with a little context. Without that context nobody could understand it. The context contains:

• the name of your family space, your time zone and currency format, today’s date and your country if it is not Germany
• the people in your family with their names and relationships, and whether each one is a child
• your lists, and which of them belongs to the sender
• the names of your appointment categories, so that moj can pick one of them
• the names of the calendars you subscribe to and who goes there, so that a note from nursery ends up with the right child
• your children’s timetables with subjects and times, if you have entered any, so that moj can tell when there is school
• your open and upcoming entries with their title, time and the people who are responsible or going
• the last few messages from the same sender

An entry marked “Only me” stays with its person. It goes into that one member’s context and nobody else’s. In the same way, an entry or a list for individual people appears only in the context of those who can see it.

moj could not work without this information. Without the people it would not know who “Mo” is. Without the open entries “Bins are out” would have nothing to tick off. Without today’s date “Thursday” would mean nothing.

Then there is whatever you send yourself. A photo or a PDF is read. A voice message is turned into text. Nothing you do not send leaves your family space.

What stays with the provider?

No history of your messages is kept there. We send every request with the instruction not to store it. OpenAI works for us as a data processor under contract. According to its own documentation, OpenAI does not use requests made through the API for training. To detect abuse, OpenAI keeps logs for up to 30 days afterwards. We checked this on 7 October 2026.

Processing at OpenAI is not limited to the EU. OpenAI does offer processing that stays within Europe, but moj does not use it yet.

None of this builds a profile of you. The AI gets the context afresh each time and remembers nothing next time.

Everything is stored with us. Our database is in Frankfurt. Your files are in private storage restricted to the EU. What that means in detail, who else is involved and how long anything is kept is set out in full in the privacy notice. What you read here is the short answer. The binding one is there.

How sure does moj have to be before it files something?

With each assessment the AI says how sure it is. moj then decides in three steps:

• Sure. moj creates the entry at once and says how to take it back.
• Unsure. moj asks exactly one short question.
• No idea. moj says so and creates nothing.

If the first answer is not sure enough, moj also asks a second, stronger model and takes the better of the two answers.

Anything still wrong after that can be put right with a sentence or the word “undo”: when moj gets it wrong.

What happens when there is a fault or a misunderstanding?

Nothing is lost in either case.

The first case is a fault. moj tells you that it did not work just now, and you simply send the message again.

The second case is the daily limit. moj asks the AI at most 300 times a day for each family. That is an emergency brake in case something gets caught in an endless loop. It is not there to save money. Once the limit is reached, moj keeps your text as a note without asking the AI. A text from a child or a caregiver goes into the history. You can find it there and send it again the next day. moj also answers no more questions until the next day.

Why AI at all?

Because the alternative would be a form. Being able to say “Dentist for Mo on Thursday at 3 pm” in one go is the whole difference from an app where you first pick the kind of entry, then the date, then the time, then the person.

That is why an AI is involved, and that comes at a price. The sections above set out exactly what the price is.

What does moj not do with AI?

Short answers

As of 12 October 2026. What moj makes of a message is shown on what you can send. Its limits are listed on what moj cannot do. The binding information is in the privacy notice.