NotForget Privacy Policy
In one paragraph
NotForget (拾藏) only keeps the links and public previews you share into it. It does not download original videos or images and never signs in to any social platform. Your saves live on your phone by default. Sign-in is Google only. Only if you subscribe to Pro and use sync are the text parts of your saves stored in our cloud database in Tokyo, Japan; Pro subscribers can also sign in to the web version with Google to view that synced data (view only, no changes); the web version uses no cookies and does not store your saves in the browser. Both AI features need your agreement first: AI lists pass text such as a save's title, caption, author and platform through our server to Anthropic (on the free plan only when you tap "Organise into lists"; on Pro, while "Auto-organise new saves" is on, new saves are sent one by one whenever the app is open in the foreground), and AI search sends only the sentence you typed, when you tap it, and no save content. No ads, no analytics or tracking tools, and we never sell your data.
Shared collections: only a collection you choose to publish becomes visible, after AI review, to people who have the link. What is public is limited to the collection name, the links, titles, ratings and permitted official thumbnails of the saves you pick, notes you choose to make public, and the nickname and preset avatar you set; a random public author code is also provided with your shares (the same for all of them, so others can block or report). Never your Google name, email or profile picture. See section 20.
1. Who is responsible for your data
NotForget (拾藏, "we") is developed and operated by the individual developer HSU CHIHHUA (publishing as Lakkev), who is responsible for handling your personal data under this policy.
Contact: privacy@lakkev.com
2. Scope
This policy covers the Android app "拾藏 NotForget" (package com.lakkev.notforget) and this website, notforget.lakkev.com (including the web version at addresses starting with /app/, see 19.1), whether or not you sign in.
The original platforms you open from NotForget (YouTube, Instagram, Douyin and so on) have their own privacy policies, which this policy does not cover.
3. Data stored on your phone
When you use NotForget, the following is stored in the app's private storage (other apps cannot read it):
- the links you share or paste, and the original share text that came with them;
- each link's public preview: title, caption, author name and handle, duration, cover image URL;
- your collections, tags, notes, ratings, watched state, and reminder times and repeats, and the automatically built list of creators;
- a small copy of the platform thumbnail (longest side at most 320 px), used only to show your own wall;
- screenshots or photos you choose as covers (longest side at most 1080 px);
- if you use AI: each save's AI results (a one-line key point, list items' names, area or address text and ingredients, and the "been there / made it" and deleted marks you set) and suggested tags, plus whether you agreed to the AI data disclosure and the "Auto-organise new saves" switch;
- if you sign in: your sign-in state (sign-in credentials and the account data in section 4);
- Pro link health check records: each save's last check time, how many "not found" answers in a row, and the "original gone" mark (see 8.4);
- app settings such as theme, the enhanced-preview switch, the clipboard-detection switch and reminder settings (including the weekend list reminder and link health check switches), and your recent search terms;
- if you use sharing: the nickname and avatar you set, the version of the sharing rules you agreed to, each shared collection's sharing settings (which saves are shared, which notes are public), what you last shared and its link, withdrawals not yet sent (removals, hidden notes, stopped shares), and authors you blocked while signed out (see section 20). This is not part of Pro sync and is not included in the app's Backup export file, but it is included in Android's Google backup (see section 12).
Purpose: letting you organise, search and find your saves. Kept until you delete it in the app, clear the app's data or uninstall. After you delete a single save, its AI results, suggested tags and cover files are deleted the next time you open the app at the latest; the small copy of the platform thumbnail is currently not removed with a single save and stays until you clear the app's data or uninstall. If you delete your account and choose "Also erase them from this phone", saves, small thumbnails, cover files and AI results are all erased from this phone; even if you keep the saves on your phone, that account's sharing settings and unsent withdrawals are erased. None of this is sent to us by default; only Pro cloud sync, AI lists, AI content reports you choose to send (see 6.4) and shares you choose to publish (see section 20) send parts of it, as described below.
4. Account and Google sign-in
The first time you open NotForget it asks you to sign in with Google; you can tap "Not now". Every on-device feature works without signing in; sync, subscriptions, AI and the web version need an account.
When you sign in, Android obtains an identity token (ID token) from Google, which our authentication service, Supabase Auth, verifies before creating your NotForget account. Supabase Auth stores the following from that token, so we receive:
- your Google account's unique identifier, email address (and whether Google has verified it), name, and the profile picture URL Google provides (the app does not use the picture); for a Google Workspace (work or school) account, also its domain;
- a random account ID we generate for you;
- when the account was created, when you last signed in, and sign-in records (time, IP address and the device's connection identifier, kept in Supabase Auth's security log; see section 15 for how long);
- a sign-in session for each device (and for each browser where you sign in to the web version), including the IP address and the device's or browser's connection identifier, used to keep you signed in and kept until you sign out, the session expires, or the account is deleted.
We never receive your Google password and we do not read your Gmail, Drive, contacts or any other Google service. Your sign-in state (sign-in credentials and the account data above) is also stored on your phone to keep you signed in and to show which account you are signed in with.
Signing in to the web version: when you sign in with Google on the web version, your browser goes first to Supabase Auth, then to Google's sign-in page, and after you sign in it returns to the web version with a one-time code (the OAuth authorisation code flow with PKCE). Supabase Auth receives the same Google account data as when you sign in to the app (the list above) and creates a sign-in session for that browser. What you enter on Google's sign-in page is handled by Google under its privacy policy.
What the web version stores in your browser: the web version keeps your sign-in state (sign-in credentials and the account data above, such as your email and account ID) in that browser's local storage (localStorage), so you do not have to sign in again next time; during sign-in it briefly keeps a one-time PKCE verifier; and only if you switch the page's language does it also remember your choice. The web version uses no cookies. Your saves, collections, notes and other content are loaded only into the memory of the open page and are never written to browser storage. "Sign out" removes the sign-in state from that browser and ends only that browser's session; the app on your phone stays signed in. If you close the tab without signing out, that browser stays signed in; after using a shared or public computer, remember to sign out.
If you are signed in but not on Pro, your saves and AI results still stay only on your phone. The cloud holds only your account, a profile (creation time and AI consent time), your subscription status, your usage counts if you use AI, any AI content reports you send (see 6.4), and data from sharing features you use (such as reports you send, your block list, save records, and shares you published earlier with Pro; see section 20).
If you do not provide this data: everything on the device keeps working, but you cannot sync, subscribe to Pro, or use AI or the web version.
5. Pro cloud sync
With Pro (including the free trial) and while signed in, NotForget syncs the following to our cloud database so you can switch phones or use several devices:
| Synced | Never synced (stays on the phone) |
|---|---|
|
|
When it syncs: when the app opens or returns to the foreground, about 5 seconds after you change a save, and when you tap "Sync now". Sync runs only while the app is running; nothing is scheduled in the background.
Where: Supabase's Tokyo, Japan region (Amazon Web Services ap-northeast-1). Traffic between the app and our servers is always encrypted (HTTPS). Row-level security makes sure each account can only read and write its own rows.
Server functions that handle requests (Supabase Edge Functions) may run in the Supabase region closest to you, but the database is only in Tokyo.
Viewing on the web: while you have an active Pro subscription, you can sign in to the web version with the same Google account and view the synced data in the table above (saves, collections, AI lists and so on). The web version reads the same database directly over HTTPS, under the same row-level security, so it can read only your own data; it is view only and cannot add, change or delete anything. Without an active Pro subscription, the web version reads only your subscription status and shows a notice.
When a subscription ends, sync stops and the web version can no longer show your saves, but cloud data is not deleted automatically; it is kept until you delete your account, and subscribing again resumes sync. The saves on your phone are not affected. You can delete your account at any time to delete the cloud data.
6. AI lists and natural-language search
NotForget has two AI features, and both need an account:
- AI lists: finds places and shops (name and area or address) and recipes (dish name and ingredients) in a save's text and turns them into lists, along with a one-line key point and suggested tags.
- Natural-language search ("AI search"): turns a search you describe in one sentence (for example, "cafés in Taipei I saved last month") into search conditions.
Before the first use the app explains what is sent and asks for your agreement; if you decline, nothing is sent. This one agreement covers both features. You can turn AI off at any time in Settings › NotForget Pro; turning it off also withdraws the consent record on our server, and both features (including auto-organising) stop sending data; your other devices ask again before their next send and never restore the consent on their own. Consent belongs to the account: signing in with a different account on the same phone asks again.
Usage limits: the two features are counted separately. Days and months follow Taiwan time (Asia/Taipei); once a limit is used up, you can use that feature again the next day or month.
| Plan | AI lists | AI search |
|---|---|---|
| Free (signed in) | 10 saves a month | 20 searches a month |
| Pro | 50 saves a day | 50 searches a day |
| During the Pro free trial | 20 saves a day | 20 searches a day |
6.1 AI lists: when data is sent
- Free (signed in): data is sent only when you open a save and tap "Organise into lists" (or organise it again), and only for that one save.
- Pro (including the trial): besides organising by hand, while "Auto-organise new saves" is on (on by default; you can turn it off in Settings › NotForget Pro), the app automatically sends new saves one by one while it is open in the foreground, covering saves made after the setting was turned on; once the day's limit is used up, it waits until the next day. The "Organise older saves" button on the Lists page sends saves that have not been organised yet, newest first, until the day's limit is used up, you tap Stop, or the app leaves the foreground. Automatic sends only ever use Pro's daily limit: once a subscription has ended, the server refuses automatic sends instead of charging the free monthly allowance.
- Nothing is sent while the app is closed or in the background, and auto-organising sends nothing unless you have agreed to the AI data disclosure.
What the app sends to our server for each save:
- platform name (for example, youtube) and kind (for example, video or post)
- author name (the author's handle when there is no name)
- title (at most 300 characters)
- caption (at most 1,500 characters)
Before sending, the app removes links and email addresses from the title, caption and author name on your phone; our server also strips everything after ? from any remaining link. Our server passes to Anthropic at most the platform, author name, title and caption; the kind is not passed on.
Never sent: the save's link, your notes, the original share text, thumbnails or any image. Anthropic never receives your name, email or account ID. We cannot analyse video frames and we do not read subtitles.
Flow: app → our Supabase server function (checks account, consent record, plan and usage count, and region) → Anthropic, PBC (USA) Claude API produces a one-line key point, list items (places: name and area or address; recipes: dish name and ingredients) and suggested tags → the result returns to your phone.
Where the result is kept: the key point and list items are stored on your phone; you can mark items as "been there / made it" or delete items that were picked up wrongly. With Pro and sync, the key point, list items and these marks are synced to our cloud database (Tokyo). A deleted item is only marked as deleted (and that mark is synced too) so it does not come back when the save is organised again; when you delete the save, its key point and list items are deleted with it. A suggested tag is added to the save only when you tap to accept it, and accepted tags sync with Pro; suggested tags you have not accepted or have skipped stay on your phone only.
Our server does not keep the text sent or the AI output and does not log them; it keeps only daily and monthly usage counts for the limits (except AI results you choose to report, see 6.4). The synced results described above are uploaded separately by your phone when it syncs.
6.2 Natural-language search
Typing in the search screen searches only on your phone and sends nothing. Only when you tap "AI search" does the app send the sentence you typed (at most 200 characters) through our server to Anthropic, which turns it into search conditions: keywords, places, kind, platform, list type (place or recipe), a saved-date range and watched state. Our server adds today's date so that phrases such as "last month" can be turned into dates.
Once the conditions return to your phone, matching them against your saves happens entirely on the phone; no save content is sent for search. Our server keeps neither the sentence you typed nor the conditions the AI returns, only a usage count (except conditions you choose to report, see 6.4). Please do not put names, phone numbers or other personal data in your search sentence.
Search on the web version is keyword-only and runs in your browser; it does not use AI and does not send the words you type.
6.3 Opening places in Google Maps
When you tap a place in a list, the app opens an ordinary Google Maps search address (https://www.google.com/maps/search/?api=1&query=…, containing that place's name and area or address text) in your browser or the Google Maps app. NotForget does not use the Google Maps API and we send no data for this; just as when you open an original post from NotForget, Google receives that search text and your connection data when you open it, and handles them under Google's privacy policy. "Search in Google Maps" on the web version works the same way and opens the same kind of address in a new tab.
6.4 How Anthropic handles data, and other notes
How Anthropic handles it: under Anthropic's Commercial Terms of Service (version effective 17 June 2025), Anthropic may not train models on customer content from its services, and processes data under its Data Processing Addendum. According to Anthropic's Privacy Center (updated 1 July 2026), API inputs and outputs are automatically deleted from its back end within 30 days of receipt or generation; if its automated safety systems flag them as a possible Usage Policy violation, inputs and outputs may be kept for up to 2 years and safety classification scores for up to 7 years, and longer where the law requires. We have no Zero Data Retention agreement with Anthropic.
Hong Kong and Macau: Anthropic does not support these regions. AI features are not offered when our server sees your connection coming from Hong Kong or Macau, or when your Google Play account country (or, if unavailable, your phone's region setting) is Hong Kong or Macau. The same applies to the AI review before publishing, so shares cannot be published or updated in these regions (see section 20).
AI review for sharing: publishing or updating a shared collection also sends the content meant to become public to Anthropic for checking. What is sent differs from the two features above (it includes the saves' URLs and the notes you chose to make public); see 20.3. Anthropic handles it as described in this section.
Reporting AI output: "Report this AI content" sits next to AI list results and AI search conditions (in "AI lists", long-press an item), so you can report without leaving the app. You pick a reason (incorrect, inappropriate, other), may add a short note (up to 300 characters) and tap "Send report". Our server then keeps: your account ID, the reason, the note, the time, the save's ID (none for AI search) and the AI result that was shown (on the detail sheet the key point, list items and the suggested tags on screen; in "AI lists" the item you long-pressed; or the search conditions the AI first read, including any you later removed, whose keywords and places come from your sentence). It does not receive the save's title, caption or author, or the whole sentence you typed. Reports do not go to Anthropic and are used only to review and improve the AI's results; the app cannot read them back, only we can see them in our back end. Each account can send up to 20 reports per 24 hours. Reports are deleted by the daily clean-up 180 days after they are sent, and with your account. For anything else you can still email notforget@lakkev.com.
AI output can be wrong (for example, a place may be wrong or closed, or a recipe's ingredients incomplete); the original work is what counts.
7. Subscriptions and payment
Pro is purchased through Google Play. Payment details (cards, carrier billing and so on) are handled by Google; we never see or store them.
We use RevenueCat, Inc. (USA) to manage subscription status and the payment records of 「Buy me a coffee」 tips. RevenueCat receives your NotForget account ID (random), your Google Play purchase records (plan, purchase and expiry times, purchase token, whether it is a trial), the store country, currency and locale, and technical data such as your IP address and the app and Android versions. If you open the plans page without signing in to see prices, or leave a tip while signed out, RevenueCat receives only a random anonymous ID (and that tip's purchase record); subscribing to Pro always requires signing in. We do not send your name or email to RevenueCat, we have turned off RevenueCat's automatic device-identifier collection, and we connect no advertising or attribution services.
RevenueCat notifies our server of your subscription status; we keep only: Pro or not, trial or not, expiry time, plan code, store name, purchase environment (production or test) and the ID of the last notification processed.
8. Fetching public previews
To show a title and cover, NotForget fetches the link's public preview from your phone (not from our servers), from the platform the link belongs to. Previews are fetched only when you share, paste or add a link, or open a save whose thumbnail needs repairing, and only for that one link. In addition, when your wall shows a save that has no small thumbnail on the phone yet (for example after syncing to a new phone or importing a backup), the app loads the cover image directly from that platform's image server (also with NotForget's identifier), only to display it, without saving it as a file. NotForget never fetches previews in the background; the only background network use is the Pro link health check (see 8.4), which only asks the official embed information whether a work still exists and fetches no preview or cover image.
8.1 Official previews and ordinary web pages
- official embed information (oEmbed) from YouTube, X and Spotify, and Apple Podcasts' public lookup;
- expanding short links (following redirects to the real address);
- public preview tags (OpenGraph) of ordinary web pages, including bilibili, public Telegram channels, news and articles;
- X posts: when the official embed has no thumbnail, the preview image from that post's public page (X's own image domain only, never the author's avatar). Like everything in this section, this does not depend on the "Enhanced previews" switch;
- cover images: downloaded once and shrunk to a small copy on your phone.
8.2 Enhanced preview (can be turned off; on by default)
Some platforms offer no public preview interface. With "Enhanced previews" on:
- Instagram: NotForget queries Instagram's public embed endpoint (
instagram.com/api/v1/oembed, which is not an interface Instagram officially offers to developers) for that post's title, author and cover URL. - Douyin, Xiaohongshu, Facebook, Threads, TikTok: NotForget loads that one work's public page once in an invisible private web view on your phone and reads the title, caption, author, cover image URL, duration and work ID.
- no sign-in, and never your accounts or cookies on these platforms;
- the web view's cookies, storage and cache are cleared before and after every load;
- Douyin only: its site shows content only after setting a cookie, so during that single load the web view accepts Douyin's own first-party cookies; they are cleared before and after, never kept and never shared with other platforms;
- known ad and tracking requests are blocked using a built-in list (including tracking subdomains and URL paths), and style sheets, fonts and audio/video are not loaded;
- the load is abandoned if it has not finished after about 7 seconds (about 12 for Douyin);
- Google Safe Browsing (built into Android WebView) checks the page.
After enhanced preview is first used, the app shows a one-time explanation and lets you turn it off. You can also turn it off at any time in Settings › Saving › Enhanced previews. Then these platforms use only the share text and a generated cover; the official previews in 8.1 keep working and your saves are not affected.
8.3 What the platforms receive
Just as if you opened the link yourself, the platform receives your IP address, the link (or, when the wall loads a cover image directly, that image's address), NotForget's identifier (User-Agent) NotForget/1.0 (+https://notforget.lakkev.com/bot) (the private web view appends it to the phone's own WebView identifier), and, during a web-view load, whatever that page's own scripts collect. They handle it under their own privacy policies. The sharer's tracking parameters are removed from the address before loading. See About NotForget's fetcher.
8.4 Link health check (Pro, on by default, can be turned off)
With Pro (on by default), once a week in the background (only while the phone is on Wi‑Fi and the battery is not low), NotForget asks the official embed information (oEmbed) of YouTube (videos), X (posts) and Spotify (podcasts) whether your saves still exist: one request per save, about 1 second apart, at most 200 per run. Before asking YouTube or X, it first asks about one public work known to exist, to confirm the platform answers normally.
- Requests go straight from your phone; your saved links never pass through our servers. The platform receives the same as in 8.3 (your IP address, that work's link and NotForget's identifier).
- The weekly check marks a save "original gone" only after two "not found" answers (HTTP 404 or 410) with no "found" in between, each at least 5 days after the previous check; the mark is cleared if it is found again. You can also clear it yourself with "It's still there" on the save or in the Originals gone list; it is then marked again only after two more "not found" answers.
- In addition, with or without Pro, when you save a link or its thumbnail is repaired, a clear "not found" from YouTube (videos), X or Spotify marks it right away; the weekly check clears the mark if the work is found.
- Every other answer (such as a sign-in requirement, a region restriction, a redirect, no connection or a busy platform) is ignored and changes nothing; when most of one platform's answers in a run are "not found", that platform's answers in that run are ignored too. Some platforms also answer "not found" for private works, so a private work may be marked. A run with no clear answer at all (for example offline) is tried again at most once, about a day later.
- At most one summary notification a week. The "original gone" mark syncs with Pro; the check records stay on the phone.
- Other platforms, and YouTube playlists and channels, are not checked. Turn it off in Settings › Reminders › Link health check; it also stops when Pro ends.
9. Clipboard
Apps such as Douyin and Xiaohongshu only offer "copy link", so NotForget can save from the clipboard. NotForget reads the clipboard only in these cases:
- when you tap the "Save to NotForget" Quick Settings tile, the "Save clipboard" home-screen widget or the persistent notification button (off by default); only NotForget itself can open these entry points, so other apps cannot trigger a read for you;
- when you tap "Paste from clipboard" on the app's "New save" screen;
- when you open or return to the app, if the system marks the clipboard content as a URL (for example, Android 12+'s automatic classification), NotForget reads it to offer saving it at the top; when the system cannot tell, nothing is read. You can turn this detection off in Settings › Saving › Detect copied links on open (on by default).
On Android 12+ the system shows a "NotForget pasted from your clipboard" notice. Clipboard content is processed only on your phone and never sent to us; only a link you choose to save becomes a save.
10. Photos and screenshots
Choosing a cover uses Android's system photo picker, so NotForget receives only the one photo you pick and needs no gallery permission. Screenshots you share in are copied into the app's private storage. These images stay on your phone: never synced or uploaded, and not included in Google cloud backup.
11. Notifications
NotForget uses notifications for watch-later reminders, the daily review, long-unwatched saves, the Pro weekend list reminder and "Originals gone" summary and, if you turn it on, the persistent clipboard notification. All reminders are scheduled on the phone; no push server is used, and the once-a-day background tidy-up only works out reminders on the phone without going online (for the link health check see 8.4). Notification permission is requested only when you first set a reminder, turn on any reminder in the reminder settings, or turn on the persistent notification; without it the link health check still updates its marks but cannot tell you. Reminder times and repeats sync with Pro; the weekend list reminder settings stay on the phone.
12. Android backup and export
- If you turn on Google backup in Android settings, NotForget's text data (the saves database, including original share text; app settings; AI results; sign-in state) is backed up to your own Google account, which we cannot access. Thumbnails, your covers, shared-in screenshots and web-view data are not backed up to the cloud.
- On Android 12+, a direct device-to-device transfer (not via the cloud) also carries thumbnails and your covers to the new phone; on Android 11 and lower it does not.
- You can also export a JSON file from Backup in the app and keep it wherever you choose.
- The sharing settings on your phone (nickname and avatar, the version of the sharing rules you agreed to, which saves are shared and which notes are public, what you last shared and its link, withdrawals not yet sent, and authors you blocked while signed out) live in the app's database and settings, so they are included in the Google backup and device-to-device transfer above, but not in the JSON export file. After restoring to a new phone, unsent withdrawals are sent only once you sign in to the same account.
13. What we never do
- no ads and no advertising ID;
- no analytics, crash-reporting or tracking tools;
- we never sell or rent your personal data or share it for advertising;
- we never download original videos, music or images or remove watermarks;
- we never ask for or keep your accounts, passwords or cookies on any social platform;
- we never read your phone's location (GPS and the like), contacts, messages or other apps' data; only when you use AI (including the AI review before publishing a share) do we work out your country or region from your connection's IP address (to check that the service is offered there), and we discard it straight away.
14. Service providers and locations
We share data only with providers needed to run NotForget, who process it on our instructions (Google Play and Google sign-in operate under Google's own terms).
| Provider | Purpose | Data | Location |
|---|---|---|---|
| Supabase, Inc. | Authentication (app and web version sign-in), Pro sync database (also read by the web version), server functions | Account data and sign-in records, synced data, subscription status, AI consent time and usage counts, AI content reports you send; AI request content passes through but is not kept. Also published shares and their management records, content under review (held briefly, see 20.3), reports, blocks and save records | Japan (Tokyo); server functions may run in other regions |
| Anthropic, PBC | AI lists and natural-language search (with your consent; lists are sent when you tap "Organise into lists" or "Organise older saves", or with auto-organise on; search only when you tap "AI search"); also the AI review before publishing a share (sent only when you tap publish or update) | Lists: platform, author, title, caption; search: the sentence you typed; share review: the collection name, nickname, titles, URLs, public notes and official thumbnail URLs meant to become public | USA |
| RevenueCat, Inc. | Subscription status and tip payment records | Account ID (or an anonymous ID when signed out), purchase records, store country, currency and locale, IP address | USA |
| Google LLC | Google sign-in (app and web version), Google Play billing, Android backup (your own account), WebView Safe Browsing, our inbox (Gmail) | Depends on the feature you use; email you send us | Per Google's services |
| Cloudflare, Inc. | Hosting this website (including the web version), forwarding email (privacy@ and notforget@lakkev.com); also generating shared-collection web pages | Connection data when you visit this website, including the web version (IP address etc.); email you send us; public content read to generate a shared page (not kept) | Global |
The platforms your links belong to also receive requests directly from your phone when you act (section 8); with the Pro link health check on, YouTube, X and Spotify also receive weekly background queries directly from your phone (section 8.4); and when you tap a place in a list, Google receives that place's search text (section 6.3). Shares you choose to publish can be seen by anyone with the link, and when a shared page is viewed, the viewer's browser or app loads thumbnails directly from the image servers of YouTube, X, Spotify and Apple (section 20); when you use the web version, your browser also loads your saves' covers directly from those four companies' image servers (see 19.1). We do not give your data to anyone else unless the law requires it (for example, a lawful court or prosecutor's order).
Your data may therefore be processed outside Taiwan (Japan, USA).
15. Retention and deletion
| Data | Kept until |
|---|---|
| Saves and images on your phone | You delete them in the app (a single delete takes effect after the "Undo" window; its cover files and AI results are deleted the next time you open the app at the latest, except the small platform thumbnail, see section 3), clear the app's data or uninstall |
| Cloud sync data (including AI list results) | You delete that save or your account; kept even after Pro ends. A deleted save is first marked as deleted in the cloud (so other devices can sync), and its key point and list items are marked as deleted with it; this content is permanently removed by the daily clean-up once the marker is 180 days old. A single list item you delete is only marked as deleted and kept (so it does not come back when the save is organised again) until you delete that save or your account |
| Account, profile, subscription status | You delete your account |
| AI content reports (reason, note, the AI result that was shown) | Removed by the daily clean-up 180 days after they are sent; deleted with your account |
| Sign-in sessions (incl. IP and device or browser user-agent) | Until you sign out, the session expires, or the account is deleted |
| The web version's sign-in state and language choice in your browser | Until you sign out of the web version (except the language choice) or clear this website's data in your browser; once the account is deleted or the session expires, the stored sign-in state can no longer be used |
| Supabase system logs (authentication audit log; API, database and server-function (Edge Functions) logs; including IP addresses) | Under the Supabase plan we currently use: authentication audit log 1 hour; API, database and server-function logs 1 day |
| AI usage counts (daily and monthly counts for lists and search, no content) | Removed by the daily clean-up 60 days after the period starts (the day itself for daily counts, the 1st of the month for monthly counts) |
| Subscription notification event IDs (no personal data; used to avoid processing twice) | 90 days |
| AI request content at Anthropic | Normally deleted within 30 days under Anthropic's rules (exceptions in section 6) |
| Published share content (collection name, nickname, avatar, and each item's link, title, rating, official thumbnail URL and public note) | Until you remove that item, hide that note, stop sharing or delete your account; when auto-hidden, taken down or when sharing is suspended it is no longer shown publicly, but is kept until then for review |
| Share management records (collection name, nickname, avatar, link code, version, the version and time of your latest agreement to the sharing rules) and your public author code | Until you delete your account; when you stop sharing, the public items and notes are deleted at once and only these management records remain (nobody else can read them), so you can manage your shares on a new phone |
| Share content under review | Cleared as soon as review passes, fails, errors or you withdraw; anything unfinished is cleared by the next daily clean-up (within about 25 hours when the schedule runs normally) |
| Review records (status, version, terms version and time agreed, no content) and daily publish counts | Removed by the daily clean-up 60 days after they are created |
| Publication settlement records (publication IDs only, no content; a publish whose result the app did not receive is settled first, so it cannot take effect after you withdraw) | Removed by the daily clean-up 60 days after they are created; deleted with your account |
| Sharing suspension flag (whether your sharing has been suspended, see 20.7) | Until you delete your account |
| Withdrawal records (used so the same withdrawal is not processed twice; no content) | Until that collection's share record or your account is deleted |
| Reports (reason, note, target, time, time handled) | Removed by the daily clean-up 180 days after they are sent (handled or not); deleted earlier if the reporter deletes their account or the reported share record is deleted |
| Save records (which account saved which shared collection on which day) | Removed by the daily clean-up after 60 days; the running save count on the collection (without who saved) is kept until that share record is deleted |
| Block list | Until you unblock, or until you or the other person deletes their account; blocks made while signed out are kept only on your phone |
| Server database backups | The Supabase free plan we currently use includes no automatic database backups, and we make no backups of our own; if we move to a plan with backups, we will update this table first |
Deleting your account deletes all data about you in our cloud database and your sign-in account (this includes all your shares and their management records, your public author code, review and withdrawal records, consent records, publish counts, reports you sent, your block list and save records; your shared pages stop working at once), and deletes your subscription record at RevenueCat; order records Google Play keeps itself are handled under Google's policies. Deleting your account does not cancel a Google Play subscription; cancel it in Google Play first. Deleting your account also does not delete the saves on your phone; to remove those, delete them in the app, clear the app's data or uninstall. See Delete account.
16. Security
- traffic between the app and our servers is always encrypted with HTTPS, and the web version always uses HTTPS too;
- the web version runs under a strict Content Security Policy (CSP): it can run only our own code, connect only to our database, and load images only from the permitted platform addresses;
- the web version keeps your sign-in state in the browser's local storage, so anyone who can use that browser can use it too; sign out when you finish on a shared or public computer;
- when fetching previews, the app upgrades
http://links tohttps://; sites that do not support encrypted connections get no preview; - row-level security separates every account in the cloud database;
- server keys (for example, for the AI and subscription services) exist only on the server, never in the app;
- data on your phone is kept in the app's private storage.
If a personal-data breach happens, we will notify you as soon as possible and report it to the authority as Taiwan's Personal Data Protection Act requires.
17. Age requirement
NotForget is for users 18 and over and is not directed at children or teenagers. If we learn an account belongs to someone under 18, we delete the account and its cloud data.
18. Your rights (Taiwan Personal Data Protection Act)
Under Article 3 of Taiwan's Personal Data Protection Act you may, for the personal data we hold about you:
- ask to check or review it;
- ask for a copy;
- ask for it to be supplemented or corrected;
- ask us to stop collecting, processing or using it;
- ask for it to be deleted.
How: data on your phone can be viewed, edited, deleted or exported as JSON directly in the app. Cloud data can be deleted with "Delete account" in the app, or by emailing privacy@lakkev.com from the Google account address you use with NotForget so we can confirm it is you. We reply within the statutory periods (15 days for review and copies; 30 days for correction, supplementing, stopping use and deletion; each may be extended once with written reasons).
Purposes (PDPA specific-purpose codes): 090 consumer and customer management and service; 135 information and communication services; 148 online shopping and other e-commerce services. Data categories: C001 identifying information (email, name, account ID; if you share, also your public nickname and public author code); C093 financial transactions (subscription status only, no card numbers); C132 unclassified data (your saves and the AI content reports you send; if you use sharing, also what you share, reports, blocks and save records). Period, location, recipients and method of use: sections 5–7, 14, 15 and 20.
You may also complain to Taiwan's personal-data protection authority.
19. This website
This website's regular pages are plain static pages: no cookies, no scripts, no analytics or tracking, no sign-in and no forms. There are two exceptions: shared-collection pages are generated dynamically and load third-party thumbnails, as described in the next paragraph; and the web version runs our own code and lets you sign in, see 19.1. The site is hosted on Cloudflare Pages; Cloudflare processes connection data (such as IP addresses) to deliver pages and protect against attacks.
Shared-collection pages (addresses starting with /c/) are generated each time someone opens them, by Cloudflare's server code (Pages Functions) reading the public content from our database, and are not cached. These pages also use no cookies, no scripts and no analytics or tracking, but they load thumbnails directly from the servers of YouTube, X, Spotify and Apple; see 20.5.
19.1 Web version
The web version (https://notforget.lakkev.com/app/) is only for Pro subscribers: after you sign in with Google, it reads the data the app has already synced to the cloud (see section 5) so you can browse your saves, collections and AI lists (places and recipes), search by keyword, and open the original links. The web version is view only: nothing can be added, changed or deleted there. If the account has no active Pro subscription, it reads only the subscription status and shows a notice.
- Code: the web version runs only our own code, with no third-party scripts, ads, analytics or tracking; it runs under a strict Content Security Policy (CSP) and asks search engines not to index it.
- Sign-in and browser data: sign-in is Google only; the flow and what is stored in your browser are described in section 4. The web version uses no cookies; the sign-in state is kept in the browser's local storage (localStorage), and your saves are loaded only into the memory of the open page, never written to the browser. As with any web page, the browser records the web version's addresses (including the words you search for) in its history, and may keep cover images it has loaded in its cache; on a shared computer, sign out and clear the browsing history when you are done. "Sign out" ends the sign-in only in that browser; the app on your phone stays signed in.
- Connections: this website (Cloudflare); our Supabase database in Tokyo (the same one as in section 5, over HTTPS, with row-level security so you can read only your own data); and Google during sign-in. Keyword search runs in your browser; nothing is sent to AI and the words you type are not sent.
- Cover images: your browser loads official thumbnails directly from the image servers of YouTube (
i.ytimg.com), X (pbs.twimg.com), Spotify (i.scdn.co) and Apple (*.mzstatic.com), the same allowlist as shared pages (see 20.5); those companies receive your IP address, browser information and the thumbnail's URL, and the page sends no referring address. Covers from other platforms are not loaded; a NotForget generated cover is drawn instead. - Leaving the web version: opening an original link takes you to that platform, whose own privacy policy applies; "Search in Google Maps" sends the place text to Google (see 6.3).
20. Shared collections
The sharing rules (content rules, reporting and enforcement) are in section 8 of the Terms.
20.1 What becomes public, and who can see it
Once you publish and it passes review, anyone who has the share link can see the following on the web or in the NotForget app, without signing in:
- the collection's name;
- each save you choose to share: its link, title, rating and a permitted official thumbnail URL (YouTube, X, Spotify and Apple Podcasts only, after review; other sources use a black-and-gold generated cover, and thumbnails on your phone or photos you picked are never uploaded);
- notes you choose to make public. Notes are hidden by default, including for saves added later; hidden notes are not sent and not stored on the server;
- the nickname and avatar you set, and your public author code (see 20.2).
Never made public or sent for sharing: your Google name, email, profile picture and account ID, your tags, watched state, reminders, original share text, screenshots and covers you picked, and any saves or notes you did not choose to share.
Shared pages are not listed or searchable anywhere in NotForget, and the web pages ask search engines not to index them; the link contains a long random code that is hard to guess. Links can still be forwarded, and anyone who gets one can see the page. Collections are private by default.
20.2 Nickname, avatar and public author code
- Before you share for the first time, the app asks you to set a nickname and pick an avatar from preset icons and colours. The app never uses your Google name as your nickname automatically, and you cannot upload your own photo.
- Like the collection name, your nickname goes through the AI review and is fixed into each published version; if you change it later, it appears only when you next publish or update that collection.
- We generate a random public author code for you (different from your account ID, and unchanged when you change your nickname). It is provided with your shared content so that others can block or report your shares; as a result, every collection you share carries the same code, and someone with two of your links can tell they come from the same author. Web pages do not display the code, but anyone with the link can obtain it.
20.3 Data sent when you publish, and the AI review
Only when you tap "Publish" or "Update share" does the app send that collection's content meant to become public (the collection name, nickname and avatar, and each chosen item's link, title, rating, official thumbnail URL and public note) to our server, together with the version of the sharing rules you agreed to. The server checks your account, Pro, region, daily publish count (currently 20 a day, Taiwan time, separate from your AI list and AI search allowance) and the content's format, and then:
- sends the collection name, nickname, and each item's title, URL and public note to Anthropic, PBC (USA) to check them against section 8.5 of the Terms. Items are labelled only by position (such as "item 3"), without your account ID or the saves' IDs;
- sends the URLs of official thumbnails that pass basic checks to Anthropic, which reads those images to check them; our server does not download or keep the images;
- makes the content public only after all of the text passes. The AI returns only fixed reason categories (such as adult content, harassment or spam); we do not keep its full reply.
During review, the content meant to become public is held temporarily on our server. It is cleared as soon as review passes, fails, errors or you withdraw; anything unfinished is cleared by the next daily clean-up at the latest (within about 25 hours when the schedule runs normally). We also keep review records without content (status, version, terms version and time agreed) and daily publish counts for 60 days. How Anthropic handles and keeps what is sent: see 6.4 (normally deleted within 30 days).
Agreement: before your first publish, the app asks you to read and tick your agreement to the sharing rules. This is separate from the agreement for AI lists; we record the version you agreed to and when. Hong Kong and Macau: as with the other AI features (see 6.4), shares cannot be published or updated in these regions.
When Pro ends, content you have shared can still be viewed until you withdraw it, but you cannot publish or change it.
20.4 Removing, hiding and stopping sharing
- When you remove a save from a share, hide notes or stop sharing, the app first stores the request on your phone (kept separately per account) and sends it automatically when you are signed in to the same account and online, without Pro and without AI. Until the server has it, the app shows "Waiting to send" and the content stays public; it shows as withdrawn only after the server confirms. If you sign out, the request stays on your phone and is sent once you sign in to that account again.
- Once the server receives it, those items or notes are deleted from the public content at once, and any content still under review is cancelled; shared pages are not cached, so the next person to open the page no longer sees them.
- When you stop sharing, the public items and notes are deleted and the link code is replaced, so the old link stops working for good; only the management records described in 20.7 remain. Sharing again later uses a new link.
- We cannot take back what people viewed, screenshotted or saved before you withdrew it.
20.5 Viewing and saving
- Web: shared pages are generated by our website and use no cookies, no scripts and no analytics or tracking. However, official thumbnails on the page are loaded by your browser directly from the image servers of YouTube, X, Spotify or Apple; those platforms receive your IP address, browser information and the thumbnail's URL, and handle them under their own privacy policies. The pages are set not to send the referring address (no-referrer).
- App: when you open a share link in the app, it uses that link to read the public content from our server; when you are signed in, the server uses your account to check whether you have blocked the author. Thumbnails are likewise loaded directly from the platforms above.
- Save (Android app only for now): copies only the link and title of the item (or whole collection) into your own saves on your phone, without the author's notes or ratings and without using the thumbnails in the shared data; the preview is then fetched again on your phone in the official ways described in section 8 (enhanced previews are not used for saves from a share). Like your other saves, these stay on your phone unless you use Pro sync. Saves you already have keep your own data.
- When you save while signed in, the server records "which account saved which collection on which day" for 60 days to count how often a collection is saved (once per person per day; the author's own saves are not counted). Saving while signed out is not recorded. This count is not shown publicly for now; if we ever use it for ranking or display, we will update this policy first.
20.6 Reporting and blocking
- Reports (sign-in needed): we keep your account ID, the reported collection, the target (the whole collection, one item or the author), the reason (adult content, harassment, spam, copyright or other), any note you add (optional, up to 300 characters), and when it was sent and handled. Authors cannot see who reported them. Each account can send up to 20 reports per 24 hours, and one person counts once per collection per review round. When 3 different users report the same collection, it is hidden automatically, and we check all open reports by hand every day. Reports are kept for 180 days.
- You can also report by emailing privacy@lakkev.com; email is handled by the mail services in section 14.
- Blocking: when you are signed in, your block list (your account ID and the blocked author's public author code) is stored on our server so you can manage it on a new phone; when signed out, it is stored only on your phone. After blocking, you no longer see that author's shares in the app, but this cannot stop them from seeing your shares through a link. Blocked authors are not notified.
20.7 Management records, suspension and account deletion
- So you can manage or withdraw your shares on a new phone, the server keeps each shared collection's management record (collection name, nickname, avatar, link code, version, the version and time of your latest agreement to the sharing rules) and withdrawal records without content; nobody else can read them.
- If a collection is taken down or your sharing is suspended, its content is no longer shown publicly but is kept for review until you withdraw it or delete your account (see Terms 8.6).
- Deleting your account deletes all your sharing data on the server (public content, management and withdrawal records, public author code, review and consent records, publish counts, reports you sent, your block list and save records), and your shared pages stop working at once; sharing settings and unsent withdrawals on your phone are also erased. Links and titles that others already saved belong to their own libraries and are not deleted.
- Retention periods are summarised in section 15.
If you do not provide this data: sharing is optional. Not using it affects nothing else; while signed out you can still view and save, but you cannot publish or report, and blocks are kept only on your phone.
21. Changes
When we change this policy we update its effective date and version; significant changes (for example a new purpose or a new provider) are announced in the app, and we ask for your consent again where needed.
Version history
- 1.7 (13 November 2026): added the web version (
/app/, Pro only, view only): new section 19.1, and the summary and sections 2, 4, 5, 6.2, 6.3, 14, 15, 16 and 19 updated to describe its Google sign-in, what it stores in your browser, its connections and how it loads cover images. - 1.6 (12 November 2026): sharing is now open; removed the "once sharing is available" notices in the summary and sections 3, 4, 6.4, 12, 13, 14, 15, 18, 19 and 20. How data is handled is unchanged.
- 1.5 (10 November 2026): new section 20, "Shared collections" (applies once sharing is available): what becomes public, visible only to people with the link, nickname and avatar, AI review by Anthropic before publishing, withdrawal, viewing and saving (web pages load thumbnails directly from each platform), reporting and blocking, retention and account deletion. The summary and sections 3, 4, 6.4, 13, 14, 15, 18 and 19 adjusted to match; former sections 20 and 21 became 21 and 22.
- 1.4 (5 November 2026): RevenueCat also handles payment records for "Buy the developer a coffee".
22. Contact
Questions about this policy or your personal data: privacy@lakkev.com
This policy is governed by the laws of the Republic of China (Taiwan). If the Chinese and English versions differ, the Chinese version prevails.