Privacy Policy
Last updated: November 6, 2026
This is an English translation of the Japanese privacy policy. If the two differ, the Japanese version prevails.
apiece (“the App”) is provided by hitohitori (“we”). The App is built so that we can remain unaware of what you read. Below we describe what the device fetches, what it sends, and what it keeps on the device.
1. Accounts, ads and analytics
There are no accounts. There are no ads. The App contains no third-party analytics or crash-reporting services (SDKs).
2. What the device fetches
| From | What is fetched | What they learn |
|---|---|---|
| Our server | The list of news sites, groups of article URLs that report the same event, and classifications. In addition, the headlines of articles provided in publishers' feeds (RSS and Atom), without altering them, together with the publisher's name and a link to the original article. We do not distribute article text or descriptions. | Your device's IP address (as with ordinary web browsing). We do not store it. |
| Each news site's (publisher's) server | Article headlines, text and images. Your device fetches them directly from the publisher. | Your device's IP address, the same as when you open that site in a browser. The publisher's handling is governed by its own policy. |
| Google Fonts | Typefaces for display. | Your device's IP address, under Google's privacy policy. |
| YouTube (Google) | Thumbnail images for video cards. When you tap play, YouTube's embedded player (privacy-enhanced mode) and captions. | Your device's IP address. When playing, the information YouTube's embedded player normally receives. Under the Google Privacy Policy; videos play under the YouTube Terms of Service. |
| Exchange-rate provider (open.er-api.com) | Only when you open the AI settings: a rate to convert your external AI balance into your country's currency. | Your device's IP address. |
3. What the device sends to our server
The device sends only the following three things. Each is sent only when you act, and only that one item.
- The feed URL of a site you added: when you add a site yourself and the “share as a candidate” setting is on, that feed's URL alone is sent once. Our server keeps only “which URL, how many times”, and does not store the sending device's IP address or the time. You can turn this sharing off at any time.
- Reports of grouping mistakes: only when you send one from “This card mixes different stories” on a card, the card's group of article URLs (up to 30) and its topic number are sent once. Our server keeps only the date received and the group of article URLs; it does not store a device identifier, IP address or time.
- Article URLs of a card you share: only when you tap “Share” on a card that combines several sites, the card's group of article URLs (up to 60) is sent once to create a share page (apiece.hitohitori.com/s/…). The headlines, site names and times on that page come from our server's own list of articles (the device sends no text). Our server keeps only the page itself and when it was created, and stores no device identifier or IP address. Anyone who knows the link can see the page, and it is deleted one year after it was created. The same group of articles always gives the same page, whoever shares it.
The articles you read, how long you read them, the topics you follow, the content of summaries, and your location are not sent. Location is not even collected.
An exception during testing only
In TestFlight and the Android test build (distributed outside the stores), the following are sent to our server only during testing. The public versions distributed through the App Store and Google Play contain the same code but do not send them.
- Records of the cards you read: the article URL, how long you looked at the card, whether you opened the original, followed it, asked for a summary, and how far you expanded it. For summaries and translations, also whether one was produced, which AI wrote it, the length of the material, and, if none was produced, the reason (a short code). The text of summaries and translations is not sent.
- Device status: the type of OS, whether the on-device AI is available, and whether an external AI is connected (the key itself is not sent).
- Sites that could not be fetched: the host name of sites whose feeds could not be fetched, the type of failure, and the count.
- Performance records: the daily reports that the iPhone compiles through MetricKit about how the App runs (CPU, memory, launch time, and records of hangs or crashes with the program location at that moment). They contain no articles read, actions or text entered.
Devices are told apart only by a random number created at installation; names, email addresses, device identifiers and IP addresses are not recorded. These records are used only to check how well grouping and ordering work. This sending stops and the records are deleted as soon as testing ends.
4. What our server sends to outside services (classifying articles and judging how stories connect)
To sort articles into genres, our server sends only the headlines and site names of publicly available articles to an external judgement service (TypeSafe AI, Inc., USA). This is limited to articles whose genre cannot be determined from the site alone, up to 3,000 a day.
Also, so that “The story so far” can reach back through stories that continue over several days, for stories reported by five or more sites our server sends two headlines, the date and the number of sites for both that story and a story from some days earlier to the same service, to judge whether one continues the other. This is up to 800 pairs a day.
- Only headlines and site names (for story connections, headlines, dates and numbers of sites) are sent. Article text and descriptions are not sent.
- Only numbers come back, such as a likelihood for each genre (for example, 0.6 for
localand 0.3 forfood) or the likelihood that one story continues another. No text is received. - No information about users (who read what, or device information) is ever included. This sending happens when articles are collected, regardless of anything users do.
- According to the recipient's published policy, the content sent is not used to train or fine-tune models and is not disclosed to third parties other than its service providers.
- We keep records of when and which articles were sent.
5. AI summaries and translation
The material for summaries, “The story so far” and “Ask AI” is the articles' headlines and descriptions, plus body text that the device fetches from the publishers' article pages. For summaries and “The story so far”, up to 600 characters of body text per site are used; for “Ask AI”, the whole article up to 12,000 characters for a single article, or up to 3,000 characters per site for stories from several sites. By default, body text is fetched for up to three sites per card (the “number of articles to read” setting changes this from 0 to 10; 0 fetches none). When writing with the on-device AI, body text is fetched only when the headline and description are not enough.
- On-device AI (Apple's foundation model on iPhone / iPad, Gemini Nano on Android): neither the material nor the summary leaves the device.
- An external AI you connect with your own key (optional): if you set an API key you obtained yourself, the material above is sent to that provider (even when the summary was written on the device, “The story so far” is written by the external AI). For “Ask AI”, your question and the earlier exchange on the same card are also sent. When the setting that automatically summarizes the cards ahead is on (it can be turned off in the AI settings), material is sent without you tapping. The recipient's handling follows that provider's policy, and any charges arise between you and that provider. The API key is stored in the device Keychain and is never sent to our server.
- Translation: the headline and text of a card in a language you don't read are translated by the device's translation feature (Apple Translation on iPhone / iPad, Google ML Kit on Android) when you tap the card's translate button (it can also be made automatic in the settings; by default, only when tapped). The headlines under “Foreign-language coverage” for stories reported by several sites are translated on the device in the same way. The text being translated does not leave the device. The first time, translation data for each language is downloaded from the OS provider.
6. What is stored on the device
The news sites and genres you chose, records of the cards you read, the topics you follow, saved summaries, behavior records (how long you looked at cards, whether you opened the original or expanded a card, and so on; used only on the device to order cards toward your preferences), and the API key (Keychain). On iPhone / iPad, the headlines of today's biggest stories are kept on the device so that Siri and iPhone search (Spotlight) can open them (they do not leave the device). Deleting the App deletes all of it. We never receive these, so there is nothing for us to delete.
7. Children
The App is not directed at children under 13.
8. Changes
When we change this policy, we announce it on this page and update the “last updated” date. Changes that add to what the device sends will be stated together with the App update.
9. Page views on this website (apiece.hitohitori.com)
To count how often each page of this website is read, we use open-source analytics software (Plausible) that runs on our own server. We use no third-party analytics service. The App itself (apiece) contains no analytics (see section 1).
- We record the URL of the page you open, the page you came from (referrer), the type of browser and OS, the type of device (desktop, tablet or phone, judged from the screen width), the country estimated from your IP address, how far you scrolled and how long the page was open, and clicks on links to other sites and on file downloads (the destination URL).
- No cookies are used and nothing is stored in your browser. IP addresses are not stored. The identifier used to group one person's visits is made from the IP address, browser information and a random value that is replaced every day (the previous day's value is deleted), so visits on different days cannot be linked to the same person.
- If your browser has “Do Not Track” turned on, nothing is recorded.
- The records are kept only on our server and are not given to third parties.
10. Provider and contact
hitohitori (hitohitori.com)
Contact: the form on the support page