Who wrote this: The Architect team. We make Architect, so read the parts about us knowing that; they match our privacy policy as of October 22, 2026. What other apps do comes from their own privacy policies and help centers, read the same day. Corrections: hello@architectapp.ai.
Can an AI journal company read your entries?
It depends on two things: where the key to your stored entries lives, and what happens when the AI writes a reply. If the company holds a key that opens your saved journal, its systems can read it, and its own rules decide which staff may. If only you hold the key, your saved entries are unreadable to the company. Either way, an AI that replies to you has to read your words at that moment.
So "can they read it" is really three questions. Can staff open my saved entries? What is sent to the AI, and what happens to it there? Who holds the key? The sections below take them one at a time, with what each company says on its own pages.
Can staff read my journal entries?
At Architect, no. Your entries are encrypted on your device before they are saved, our servers hold only encrypted text we cannot open, and there is no admin view of anyone's journal. Other apps answer differently, and the clearest ones say exactly how.
- Day One says its employees will never be able to read your encrypted entries.
- Rosebud's privacy policy says its broader staff has no access to production data, and that access to journal content is limited to a small number of authorized people, with two-factor authentication and audit logs, to fix a technical issue, answer a support request or investigate a safety or security concern.
- Life Note's privacy policy says entries decrypted for a mentor's reply are not read by its staff.
- Stoic says your journal stays on your device or in your personal iCloud, and that it does not have access to it.
- Mindsera's policy lists restricted staff access among its safeguards.
"Restricted access" describes a rule the company follows. A design in which the company never has a usable key describes a limit on what it can do. Both can be honest. They are not the same promise, and a privacy page should tell you which one it is making.
What data is sent for AI processing?
The text the AI needs to write its reply, in readable form, sent to an AI model provider. That is true of every AI journal that writes its replies on a server. What differs is how much text is sent, whether the provider keeps it, and whether it can be used to train AI models.
What Architect sends
- To write a reply: your entry and the selected context the mentor needs, such as recent entries and its dated notes, go in readable form to our AI model provider for that moment. They are not used to train AI models.
- To build the mentor's memory: the same provider reads the entry to make short dated notes, which are then encrypted the same way as your journal (how that memory works).
- While you type, with memory on: what you are writing may be sent so the mentor can find related past entries. It is never stored.
- Voice: a voice note is transcribed by a speech-recognition provider. When you tap Listen, or talk with your mentor in the Sitting, the reply's text, which can quote what you wrote, goes to a voice provider to make the audio.
- Photos: photos added since October 20, 2026 are encrypted on your device like your words; the AI sees a photo only while it writes your reply.
We keep the names of these providers off our website. A list of the service providers we use is available on request at hello@architectapp.ai.
What other apps say they send
- Day One: its AI features are opt-in. The content a feature needs is temporarily decrypted on your device and sent to an AI service; Day One says the service does not use it for training and that it does not keep AI requests on its servers once they are processed.
- Stoic: when you use an AI feature, the specific entry you are editing is sent, not your full history, and its AI provider may keep it for up to 30 days for safety and misuse prevention.
- Rosebud: its policy says entry contents are sent without user-identifying information, under zero data retention agreements.
- Mindsera: the minimum content needed for the feature you request; its policy says journal content is not used to train AI models.
- Reflection: its AI analyzes your entries through a third-party AI provider, and it says your data is never used to train AI models.
- Life Note: its policy says entries are decrypted at one point only, when an AI mentor reflects on what you wrote in order to respond.
Who holds the encryption key?
The key decides who can open your stored journal. Look for three things: where the key is created, where it is kept, and what opens your journal on a new device.
- Architect: the key is made on your device. We keep only a sealed copy that your twelve recovery words, or a private password you choose instead, can open, so we never hold it in a form we could use. If you lose the words or the password and your session ends, your encrypted history cannot be recovered, by us or anyone.
- Day One: stores the encryption key in your own iCloud or Google Drive by default, so you do not lose access to it.
- Apple Journal: Apple's support page lists Journal data among the categories whose keys are kept on your trusted devices.
- Life Note: its policy says entries are encrypted with AES-GCM (256-bit) before storage, and that the key protecting each account's key sits in a managed cloud key service backed by hardware security modules, never in readable form on its application servers.
- Rosebud, Mindsera and Reflection describe encryption in transit and at rest; Reflection names AES-256 at rest.
Can I bring my own encryption key?
Not in the strict sense, as far as the pages we read go: none of the apps above describes a way to supply a key you made elsewhere. The practical version of the question is whether the key is made on your device and opened only by something you hold. That is how Architect works, and it has a real cost. No one can reset your key for you, so keep your recovery words somewhere safe.
Why AI replies and end-to-end encryption pull against each other
End-to-end encryption means only the devices at the two ends can decrypt the data; the service in the middle stores and moves it but cannot read it. Apple's support page, for example, defines end-to-end encrypted data as data that can be decrypted only on your trusted devices.
An AI model that writes a reply is not one of your devices. It runs on a server and needs readable text, the same way a person needs to see a letter to answer it. So an AI journal that replies has three honest options, and the companies above use all three.
1. Keep the AI on your device. Apple's Journaling Suggestions use on-device processing. The trade-off is that Apple Journal does not write back to your entries.
2. Decrypt for the moment of processing. Day One's own guide to its AI features says the content a feature needs is decrypted on your device and sent to an AI service, and that end-to-end encrypted content is temporarily not end-to-end encrypted while that happens.
3. Store it encrypted, and send readable text only to write the reply. Architect and Life Note both describe this: entries are encrypted before they are saved, and the text is readable to the AI model for the moment it takes to reply.
When an app describes itself as end-to-end encrypted, ask what happens at reply time. If the text is decrypted and sent to a model, the encryption covers storage and transit, not the reply. Mindsera's policy, for instance, says plainly that it does not currently use end-to-end encryption, because its AI features need its systems and providers to process the content.
None of this makes AI journaling unsafe. It means the useful promise is specific: stored entries the company cannot open, text sent only to write the reply, not kept there, not trained on. Check each part separately, on the app's own pages.
Can I use a private AI journal in a browser?
Yes. Architect runs in any modern browser at architectapp.ai/app, on a computer or a phone, and your entries are encrypted on your device before they are saved. On a phone you can add it to your home screen, where it opens full screen like an app.
Reflection, Life Note, Mindsera, Stoic and Day One also have web versions, and Rosebud's web app is a companion to its mobile apps. Apple Journal is an app on iPhone, iPad and Mac.
In a browser the questions above stay the same, and you add three of your own. Is this a shared computer? Will you sign out when you finish? Do you trust the extensions you have installed, since some can read the pages you open? On a shared phone, a PIN helps: in Architect it is in Settings, Privacy & Security, App PIN Lock.
How Architect handles your journal, in plain words
Your journal is encrypted where it's stored, and the key stays with you.
- Stored: your entries, the mentor's replies and its notes about you are encrypted on your device with AES-256 before they are saved. There is no admin view of your journal.
- Processed: to write a reply, the text it needs is read by our AI model provider for that moment. Nothing you write is used to train AI models.
- Never sold, and never used for ad targeting.
- Yours to take or delete: export your whole journal, free on every plan (Settings, Your Data, Take a copy). Deleting your account permanently deletes it.
- The cost of holding the key: if you lose your twelve recovery words or private password and your session ends, your encrypted history cannot be recovered.
The full text is our privacy policy, and the short version is can the company read my entries? Prices and the free allowance are on Architect facts: writing is free, and you start with 11 mentor responses, no card required.
Try it free with one real thought →
Questions to ask any AI journal before you write something real
- Can staff open my saved entries, and under what rule?
- What exactly is sent to the AI to write a reply: one entry, recent entries, or notes about me?
- Does the AI provider keep it, and for how long? Is it used for training?
- Who holds the key, and what opens my journal on a new device?
- What happens to my voice recordings and photos?
- What does deleting my account remove?
Try it free with one real thought →
An app that answers all six on its own pages, in plain words, is easier to trust than one with a single reassuring line. That includes us: if something here is unclear, write to hello@architectapp.ai. For more on the trade-offs, read is it safe to tell an AI your secrets?; if you are journaling in a general chatbot today, see Architect vs ChatGPT.
Choosing on features as well as privacy? The best AI journal that writes back compares replies, memory and voice, and eight AI journaling apps are side by side on one page.