Privacy and AI

What an AI assistant sees in your to-do list

Only what the app sends it. An assistant inside a to-do app sees your request plus whatever items the app attaches or looks up for it: titles, notes, dates, tags. Some features run on your iPhone and nothing leaves it; others go to a server and on to an AI provider. Before you switch one on, find out whether it asks first, what a request contains, where it runs, what the provider keeps, and whether you can keep some items out.

Coming soon to iPhone and iPad

Last checked

On this page

The assistant only knows what the app sends

A language model does not live inside your to-do app and browse it. It sits somewhere, on your phone or on a server, and it receives text. Everything it “knows” about your list arrives in a request the app builds. So the useful question is not “can the AI see my list?” but “what does this app put in a request, and when?”

A request can hold several kinds of things:

  1. What you typed or said. Your message, or your words after dictation turns them into text.
  2. Context the app attaches. Some apps add the item you are looking at, today’s list, or a project’s tasks so the answer can be specific.
  3. What the model looks up. Many assistants can ask the app for more as they go: “everything due this week”, “the notes in this project”. The app runs that search on your phone and sends back what it found. Those results become part of the conversation too.
  4. The earlier conversation. A chat usually sends the previous messages again with each new one, so something you mentioned ten minutes ago may travel again.

This is why the same assistant can feel harmless when you ask it to reword one task and reach much further when you ask “what should I do today?”. The second question needs your list, and the app will fetch it.

Where the request goes

There are three common places a request ends up, and they make very different promises.

On your iPhone. Apple lets apps use the on-device model at the core of Apple Intelligence through its Foundation Models framework, and Apple says features built this way work offline. When a feature runs here, the text never leaves the phone. The trade-off is size: a phone-sized model handles short, well-defined jobs better than long open conversations.

Apple’s Private Cloud Compute. For Apple Intelligence tasks that need a bigger model, Apple says the request may go to Private Cloud Compute, where the data “is not stored or made accessible to Apple” and is used only to fulfil the request. This applies to Apple’s own features, not to an app that talks to a different provider.

A third-party provider, often through the app’s server. Many assistants send requests to a company such as OpenAI, either straight from the app or through the developer’s own server in between. Two policies then matter: the developer’s, for its server, and the provider’s. OpenAI’s data controls page says data sent to its API has not been used to train its models since 1 March 2023 unless the customer opts in, and that abuse monitoring logs, which can contain prompts and responses, are kept for up to 30 days by default. Customers approved for Zero Data Retention can have their content left out of those logs. Which arrangement an app has is something only the app can tell you.

Our guide to on-device and cloud AI goes further into the difference.

What a privacy label will and won’t show

The App Store privacy label is a good first look, but it was not designed to describe AI requests. Apple’s developer guidance defines “collect” as sending data off the device in a way that lets the developer or its partners access it for longer than it takes to service the request. Data that is sent to a server and discarded right after the request is serviced does not have to be declared. So an assistant whose server passes your words along without keeping them can be absent from the label, and that is consistent with the rules.

Labels are also written by the developer. Third-party App Store pages say so under the label: “This information has not been verified by Apple.” Our guide to reading to-do app privacy labels compares what well-known to-do apps declare.

Five questions to ask before you switch an assistant on

  1. Does it ask before the first request? An assistant that is on by default has already decided for you. One that asks gives you the moment to read what it does.
  2. What goes into a request? Only your message, or your list, notes and earlier chat as well? A good app says so in plain words.
  3. Where does it run? On the phone, on Apple’s Private Cloud Compute, or with a named provider. “In the cloud” is not an answer.
  4. Who keeps what, and for how long? The developer’s server and the AI provider can each keep data. Ask about both, and about training.
  5. Can you keep some things out, and undo what it changes? A way to mark items as off-limits matters more than any promise, because it does not depend on anyone’s policy. Our guide to keeping notes and to-dos away from AI covers the options on iPhone, including Apple’s own settings for Siri.

One more habit costs nothing: when you paste a list into a general chatbot, you are using that chatbot’s rules, not your to-do app’s. If a list holds health details, money or other people’s information, trim it before you paste.

Where Dozen helps, and where it doesn’t

Dozen’s Assistant was built around these questions. The Assistant is optional, and it asks first: nothing is sent to an AI provider until you allow it. Cloud requests go through Loopware’s own server to OpenAI, which passes each one on without logging or storing what you wrote. Anything you tag No AI is kept out of the Assistant’s requests. Anything tagged No AI also stays out of Spotlight, Siri and Shortcuts. Suggest steps uses your iPhone’s own model, and asks OpenAI only with Dozen Pro, after your permission, when the device cannot make the steps itself. Ask the Assistant for one thing and it is done, with Undo; ask for several and it shows you first, and nothing changes until you say. You can talk to the Assistant: your words appear in the message field and nothing is sent until you tap Send. There is more on the No AI tag and on the Assistant.

Where it does not help: The Assistant is part of Dozen Pro. The fact above covers Loopware’s own server; it says nothing about a chatbot you paste your list into yourself. Dozen is not on the App Store yet. And if you would rather not use an AI assistant at all, you never have to switch one on: a to-do list works perfectly well without one.

Questions and answers

Does an AI assistant read my whole to-do list?

Not by itself. A language model knows nothing about your list until the app sends it text. Some apps send only what you typed; others attach recent items, or let the model ask the app to look things up, and every item that comes back is part of the request. Ask the app which it does.

Is my to-do list used to train the AI?

It depends on the provider and the agreement the app has with it. OpenAI, for example, says data sent to its API has not been used to train its models since 1 March 2023 unless the customer opts in. A chatbot you use directly can have different rules, so check its own settings.

Can I see when my iPhone sends a request to Apple’s cloud?

Yes, for Apple Intelligence. Apple describes a report under Settings, Privacy & Security, Apple Intelligence & PCC Report, which logs requests sent to Private Cloud Compute for the period you choose. It does not cover requests an app sends to its own server or to another provider.

Sources

What this page says about Dozen was checked against the app on .

Last checked:

Get Dozen

Dozen is free to download. Dozen Pro, a monthly or yearly subscription, adds the Assistant.

Coming soon to iPhone and iPad

Dozen: Tasks & Notes

Dozen is free to download. Dozen Pro, a monthly or yearly subscription, adds the Assistant.

Coming soon to iPhone and iPad