A family app without AI: why Famileze uses none
In short
Famileze, the private family archive for iPhone, uses no cloud AI and no language model. Reading dates out of a school email, recognising the words in a photographed flyer and suggesting a folder are all done by small, predictable tools on your device. Your family's content never trains any model, and we couldn't use it: we cannot read it.
Why we left AI out
A family archive holds the most personal things a parent owns: photos of a newborn, immunisation records, a letter written for an eighteenth birthday. Your children's childhood is an archive, not content. It should sit somewhere it isn't rated, ranked or resurfaced.
Much of today's software sends what you type or photograph to a large AI model in the cloud, where it is read and answered. That design can be genuinely useful. It also means your words and pictures have to be readable by someone else's computer.
Famileze is built the other way around. Your photos, notes, calendar and records are encrypted on your device before they leave, so what is stored in our cloud is scrambled text that we hold no key to. A cloud AI could not read it either, and that is the point. So the clever parts have to run where your family's content is already readable: on your own phone.
What Famileze uses instead
Reading dates out of an email
Forward a school newsletter to your family's Famileze address and the app lists the dates it finds, each with the sentence it came from. This is done by a rule-based date reader that runs on your iPhone. It knows the many ways dates and times get written, including phrases like "every Tuesday" or "for six weeks". Give it the same email twice and you get the same answer twice.
It won't always be right. When it isn't, you can see why, because the source sentence sits under every date. Nothing reaches your calendar until you tap Add, and you can edit a date before you save it.
Recognising text in a photo
School news often comes as a photo: a term planner snapped off the noticeboard, a party invitation, a screenshot. In the iPhone app, the words in an emailed photo or scanned PDF are read using the iPhone's own built-in text recognition, on the phone. It returns the letters that are literally in the picture and nothing else. It doesn't describe the photo or guess at what it means. Those words are then handed to the same date reader.
Suggesting a folder
When an email arrives, Famileze suggests where to file it. The suggestion comes from a plain word list (excursion, uniform and term point to School; dentist and immunisation point to Health) and from where you filed the last email from the same sender. The folder is a dropdown, one tap from changing, and nothing is filed until you tap File.
Saying it out loud
Tap the voice button (Say it — we'll file it), say what you need to remember, and Famileze sorts it into the calendar, a to-do, a shopping list, a note or a moment. The sorting is done by a deterministic parser, and everything it worked out is shown on one review screen. Nothing is saved until you confirm.
Why small and predictable is better here
- You can check it. Every suggestion shows its working: the sentence a date came from, and a folder you can change in one tap.
- It doesn't invent. A rule-based tool finds what is written. It won't make up an event that wasn't in the email.
- You stay in charge. Every filing, calendar change and send is confirmed by you first.
- There is nothing to send away. Because the tools run on your phone, your family's content doesn't need to be readable by anyone else for them to work.
What it means for your family's privacy
- No training, ever. Your family's content is never used to train any machine-learning model, ours or anyone else's. We couldn't do so even if we changed our minds, because we cannot read it.
- Unreadable to us. Photos, captions, calendar entries, health records and letters are encrypted on your device with AES-256-GCM before upload. We hold no key and there is no back door.
- No analytics in the app. There's no analytics or tracking software in the app and no advertising, so there is no profile of your family being built.
What it doesn't do
- iPhone dictation goes through Apple. The 🎙 button on iPhone uses Apple's speech recognition, and Apple may send that audio to its servers to turn it into text. We never receive the audio. Type instead if you'd rather it stayed on the phone.
- Not everything is end-to-end encrypted. Email sent to your family address and newsletters you send out are ordinary email, and pages you choose to publish are ordinary web pages, because someone outside your family is at the other end.
- It won't write for you. Famileze doesn't caption your photos, summarise your child's year or chat with you. Your words stay yours.
- It can misread a date. That's why every suggestion waits for your approval.
Questions parents ask
- Does Famileze use ChatGPT or any other AI model?
No. There is no cloud AI and no language model anywhere in Famileze. Its smart features are small tools that run on your device.
- Will my children's photos be used to train AI?
No. Your family's content is never used to train any model, and we couldn't use it if we wanted to, because it is encrypted on your device and we cannot read it.
- How does it find dates in an email without AI?
A rule-based date reader on your iPhone looks for the ways dates and times are written, then shows each one with the sentence it came from for you to approve.
- Does anything leave my phone?
Yes, encrypted. Your archive is backed up to the cloud so both parents stay in step and a lost phone costs nothing, but it is encrypted on your device first, so we can't read it. The exceptions are iPhone dictation, which uses Apple's speech recognition, and the things that must travel as ordinary email or web pages: newsletters you send and pages you choose to publish.
Related
- How to turn school emails into family calendar events
- End-to-end encrypted family photos, explained
- Childhood is an archive, not content
- A private family photo timeline
Famileze is coming to iPhone. Join the waitlist to hear the day it launches.