StarWrite Privacy Policy

Last updated: October 8, 2026 · Effective date: the official launch date of the StarWrite service (announced separately within the service) · Version: v1.5 publication candidate

This Privacy Policy (the “Policy”) explains how Passion Factory Corporation (주식회사 열정팩토리, the “Company”) collects, uses, retains, outsources the processing of, transfers across borders, and destroys personal information through the StarWrite mobile app provided by the Company (iOS and Android, the “App”) and the StarWrite website operated by the Company (starwrite.me, the “Website”), and how you can exercise your rights.

StarWrite’s core journaling and review features are provided through the App. The Website is operated to introduce the service, provide legal documents, accept pre-launch notification sign-ups, and for similar purposes. Provisions specific to the App, such as App sign-up, entries, and AI processing, apply to App users, and provisions concerning Website visits and launch notification sign-ups apply to users of the Website.

This Policy has been prepared primarily under the Personal Information Protection Act (PIPA, Republic of Korea) and related laws. For users residing in Japan, matters under the Act on the Protection of Personal Information (APPI, Japan) additionally apply, and for users residing in California, United States, matters under the California Consumer Privacy Act (CCPA, including amendments made by the CPRA) additionally apply (Section 17).

Key terms used in this Policy are explained under “Definitions” below. Any other term relating to the App service that is not defined here has the meaning given in the StarWrite Terms of Service.

Key Points

This summary is provided to help you understand this Policy. Your specific rights and obligations are governed by the provisions in the body of this Policy.

Item

Details

Minimal collection

The Company processes only the account information needed for sign-in (Google or Apple), the entries you leave, and the minimum technical information needed to operate the Service. The App does not request camera, photo, location, contacts, microphone, or notification permissions

Sentences are stored on servers (currently in the United States)

As of the Reference Date, the sentences you leave are stored in Google Firebase (Cloud Firestore, United States). Because this transfer is essential to providing the Service, if you do not want it, you must refrain from using the Service or delete your account. Data in transit is encrypted, and within the App you can access only your own entries. The storage location and cloud provider may change depending on future infrastructure configuration; if they change, the Company will update this Policy and notify you (Sections 11 and 13)

AI reflections only with separate consent

The feature that sends your sentence to AI and returns a reflection (an AI response; some screens simply call it a “reflection”) works only if you give separate consent on the consent screen shown right after sign-in and age confirmation. If you consent, today’s sentence and the previous reflection conversation (including, where recall is provided, the sentences of the stars you picked in a weekly review), as well as, when a constellation is completed, the sentences from that period used to create the monthly comment (Section 8.10), are sent through the StarWrite server and Vercel AI Gateway (the United States and other regions operated by Vercel) to the Claude model provider (Anthropic, PBC, or a Claude hosting provider to which the gateway routes the request). The processing region is not fixed to the United States (Section 8.2). Under the same consent, your sentences are also used for crisis-signal screening (Section 7.3), and the response and the sentence are also sent to TypeSafe AI to check whether the generated response followed the conversation rules (Appendix A.4). If you do not consent, your sentences are not sent to any external AI. Even if you consent, the sentences and conversations sent are used only to generate responses and perform checks, and are not stored by the AI processing providers (Section 8). You can turn this off at any time in the top bar under “Settings” > “Privacy” > “Send my sentences to AI for reflections”. If you do not consent or you withdraw consent, your entries and stars remain as they are, and a fixed message is shown instead of a reflection

No AI training; zero data retention

What you enter is not used to train AI models. The Company applies settings that prohibit training use and require zero data retention (ZDR) to every AI request, so the sentences and conversations sent are used only to generate responses and are not stored by the AI processing providers. Operational metadata that does not contain sentence or conversation content may be recorded separately (Sections 8.5 and 9.1)

No advertising or sales

The Company currently does not display advertising, does not use any ad-serving SDK, and does not sell personal information or entries. If the Company introduces advertising or paid services in the future, it will update this Policy and obtain any consent required by law before doing so. The Company does not collect the Android Advertising ID (AAID) or the iOS Identifier for Advertisers (IDFA) for usage statistics purposes

Who can see what I write

The App has no feature that shows your entries to other users. You may yourself export a card image of a completed constellation (name, period, star shape, and monthly comment, excluding the original sentences) using your device’s share function, and the Company does not receive the exported image (Section 8.10). The Company prohibits its officers and employees from viewing entries in the ordinary course. Although server data is technically accessible, authorized personnel access it only to the minimum extent necessary and only in very limited cases, such as at your request (including reviewing reports and handling requests to exercise your rights), responding to outages or security incidents, or fulfilling legal obligations. This is an operational policy, and entries are not end-to-end encrypted. Guardians also cannot automatically view your entries (Section 18.3). If you report a reflection response, an operator may open the entry (sentence) to which that response is attached if necessary for the review (Section 8.9 and Appendix B). If you consent to AI reflections, your sentences are sent through Vercel to the Claude model provider and TypeSafe AI (Section 8)

Account deletion

You can delete your account at any time in the top bar under “Settings” > “Account” > “Delete account”. Your entries, consent records, and reports on the server, and your account, are deleted immediately and cannot be restored. The following are not deleted immediately: server backups (destroyed after their retention period expires), sign-in account information (removed from Firebase Authentication’s operational and backup systems within 180 days after deletion), server operation logs (after their 30-day retention period), and usage statistics already sent to analytics tools. Information that relevant laws require to be retained (e.g., payment and transaction records) may be retained to the minimum extent necessary for the period prescribed by law (Section 9.3)

Deleting a single sentence; signing out

At present, you cannot delete a single sentence in the App, so you must request it by email (Section 14.2). If you sign out, data stored on the server, such as entries, reflection conversations, and weekly reviews, remains and only the copy on this device is deleted; drafts in progress and sentences not yet uploaded to the server are lost and cannot be recovered. A feature for exporting entries to a file is not provided at launch

Usage statistics only with consent

Usage statistics are collected through Google Analytics and PostHog only if you consent. The default is off, and you can use all features even if you do not consent. Usage statistics do not contain original sentences, drafts in progress, or reflection conversations, but do contain classification values such as the entry’s tone (good, grateful, calm, heavy) and character-count range. You can withdraw consent at any time in the top bar under “Settings” > “Privacy” > “Send usage statistics” (Section 3.3)

Age 16 or older

The Service is available only to persons aged 16 or older. Right after Google or Apple sign-in, the App asks you to confirm that you are 16 or older; if you answer that you are under 16, you cannot use the Service and are signed out. Persons under 16 cannot sign up (Section 18)

Crisis signals and safety responses

If you have consented to AI reflections, the Service screens today’s sentence for signals of self-harm or suicide risk using keyword rules and AI classification. If a risk signal is detected, a predefined safety response containing regional support hotlines is shown instead of an AI response, and for 30 days thereafter no recall that revisits past sentences is made. This screening is not profiling that infers your tendencies but a safety measure concerning that single sentence, and there is no separate setting to turn it off (if you decline or withdraw consent to AI reflections, no screening is performed). Whatever you write, your sentence still becomes a star; no person views it in real time, and neither the Company, guardians, nor any outside organization is notified (Section 7.3)

Not a medical service

StarWrite is not a medical, counseling, or crisis-response service and does not replace professional help. If there is an urgent risk, do not rely on the Service; seek help directly from emergency services, medical institutions, or professional support organizations where you are (Section 7.4)

Contact

contact@passionfactory.ai (Passion Factory Corporation, the operator of StarWrite). Please include “StarWrite” in the subject line (Section 20)

Definitions

Term

Meaning

Entry, sentence, star

An entry is the data you leave each day, and the sentence is the body of an entry. A star is how an entry is drawn on the screen

Reflection

A short AI response in which the Universe (the narrator within the Service) reflects back the sentence you left. Some screens simply call it a “reflection”

Recall

A feature that adds to a reflection a sentence that gently revisits a past entry. Not provided as of the Reference Date (Section 8.7)

Weekly review

A feature for looking back by picking stars from a week’s entries. While you are drawing a constellation, you look back in the same way at each 7-position segment of the constellation; this is called a segment review

Constellation, monthly comment

A constellation is a set of entries, organized by month, drawn by connecting one star per day, one position at a time. A monthly comment is a one-line comment created once, based on the entries from that period, when a constellation is completed (Section 8.10)

Safety response

A predefined response containing information on support hotlines, shown instead of an AI response when a signal of self-harm or suicide risk is detected in a sentence (Section 7.3)

Tone

A classification value attached to an entry (good, grateful, calm, heavy, or unspecified)

Day boundary

The reference time, set in the App’s settings, that determines which date an entry belongs to

Outsourcing of processing, processor

Outsourcing of processing means the Company entrusting the processing of personal information to another company in order to provide the Service, and that company is called a processor

Reference Date

In this Policy, the “Reference Date” means the effective date of this Policy. Statements about infrastructure, such as processors and processing locations, describe the configuration as of the Reference Date

Cross-border transfer

The storage of personal information outside the Republic of Korea or its transfer to a company located there. For StarWrite, as of the Reference Date, this centers on the United States, where the servers and databases are located; usage statistics in Google Analytics, app integrity checks, crash reports, server operation logs, and AI processing may also be processed in regions outside the United States (Section 13)

Data subject

You, the user, as the owner of the personal information

App instance ID, Firebase Installation ID

Identifiers that the analytics tool and the Firebase SDK assign to each app installation. The Company does not link these values to your account

Cross-context behavioral advertising

Personalized advertising based on your activity on other companies’ apps and websites

Table of Contents

  • 1. Company Information
  • 2. Scope of This Policy
  • 3. Personal Information Processed
  • 4. How Personal Information Is Collected
  • 5. Purposes of Processing Personal Information
  • 6. Legal Bases for Processing Personal Information
  • 7. Entries Containing Sensitive Content
  • 8. AI Reflections and Personal Information Processing
  • 9. Retention and Use Periods
  • 10. Procedures and Methods for Destroying Personal Information
  • 11. Measures to Ensure the Security of Personal Information
  • 12. Outsourcing of Processing
  • 13. Cross-Border Transfers of Personal Information
  • 14. Data Subject Rights and How to Exercise Them
  • 15. Data Stored on Your Device and Automatic Collection Tools
  • 16. Other Grounds for Providing Personal Information
  • 17. Additional Regional Information
  • 18. Minimum Age and Protection of Children’s Personal Information
  • 19. Changes to This Policy
  • 20. Contact Information
  • Appendix A: Processors and Cross-Border Transfers in Detail
  • Appendix B: Purposes for Which the Company Does Not Use Entry Content

1. Company Information

1.1 Personal Information Controller

Item

Details

Trade name (company name)

주식회사 열정팩토리

English name

Passion Factory Corporation

Representative

Jeonghoon Lee

Role

Personal information controller (data controller)

Address

161 Hakdong-ro, Gangnam-gu, Seoul, Republic of Korea

Business registration number

114-87-18004

Email

contact@passionfactory.ai

Phone

+82-70-7756-1070

Hosting provider

Google LLC (Firebase / Google Cloud — the database and server functions run on infrastructure located in the United States; usage statistics, app integrity checks, and server operation logs run on Google’s global infrastructure (including the United States)). This is the configuration as of the Reference Date and may change depending on future infrastructure configuration. If it changes, the Company will update this Policy and notify you, and will follow the applicable procedures for any change that requires consent or notice under law

1.2 Chief Privacy Officer

Item

Details

Name

Jeonghoon Lee

Title

Chief Executive Officer (CEO)

Email

contact@passionfactory.ai (please include “StarWrite” in the subject line)

Phone

+82-70-7756-1070

1.3 Privacy Officer

Day-to-day inquiries about the processing of personal information and requests to exercise rights are received through the same contact details as those of the Chief Privacy Officer, and, as a rule, by email.

2. Scope of This Policy

  1. This Policy applies to persons who install the App and sign in with a Google or Apple account to use StarWrite’s journaling and review features, and to persons who visit the Website or sign up for launch notifications. To use the App’s journaling and review features, you must create an account by signing in with Google or Apple.
  2. StarWrite is a global service. However, EU/EEA member states and the United Kingdom are currently excluded from the official service area. For regional information, please see Section 17.
  3. If this Policy and the StarWrite Terms of Service conflict on matters concerning the processing of personal information, this Policy prevails.

3. Personal Information Processed

3.1 Information You Provide Directly

Category

Items

Storage location

Age confirmation

Time at which you confirmed that you are 16 or older

Cloud Firestore (United States) account document + cache on your device

AI processing consent record

Decision to consent or decline, time of the decision, time recorded on the server, version of the consent notice (does not include original sentences)

Cloud Firestore (United States) + cache on your device

Usage statistics consent

Decision to consent or decline, per account

Your device

Account settings

Day boundary time

Cloud Firestore (United States) account document + cache on your device

Device settings

App language

Your device

Reflection reports

Text of the reported reflection response (up to 2,000 characters), report reason (“It hurt my feelings”, “It isn’t accurate”, “It didn’t sound like StarWrite”, “Something else”), response type (standard reflection, safety response, fallback response), identifier of the relevant entry, time of the response, time of the report

Cloud Firestore (United States)

Email inquiries

Email address, content of the inquiry

Company email

Website launch notification sign-ups

Email address, language, referral source. When you sign up, the request’s IP address, browser information (User-Agent), and referring page may be stored together with the sign-up record

Cloudflare D1 (Western United States)

3.2 Content You Leave

In this Policy, “sentence” means the body of an entry.

Category

Items

Storage location

Sent to server

Entries (one sentence a day)

Sentence, time of the entry, date to which the entry belongs, tone (good, grateful, calm, heavy, unspecified)

Cloud Firestore (United States) + cache on your device

Stored on server

Weekly reviews and segment reviews

Identifiers of the picked entries, time of picking or skipping, number of times you postponed the invitation, whether recall was used (does not include sentences or tones)

Cloud Firestore (United States) + cache on your device

Stored on server

Reflection conversations

History of your messages and reflection responses, response type (standard reflection, safety response, fallback response)

Cloud Firestore (United States, stored per entry) + cache on your device

Stored on server. Used as context for the next reflection and deleted together with your account when you delete it (Sections 8.3 and 8.6)

Constellations

Constellation shape and position layout, constellation name, monthly comment created upon completion, identifier of the constellation in which an entry is placed and its position number, time at which constellation guidance was shown and number of times you declined it

Cloud Firestore (United States) + cache on your device

Stored on server (for the monthly comment, see Section 8.10)

Submission processing records

Identifier of the representative entry for each date; identifier, processing status, and time of each sentence submission request (does not include original sentences). The App cannot access these records

Cloud Firestore (United States)

Stored on server

Drafts in progress

Sentences not yet submitted, submission requests whose transmission was not completed

Your device

Device only (submissions whose transmission was not completed are sent to the server again at the next launch)

Depending on your own judgment, the sentences you leave may contain sensitive content concerning emotions, health, mental health, and the like. This is addressed separately in Section 7.

3.3 Information Collected Automatically

Category

Items

Collection method

Storage location

Usage statistics

Usage events such as screen views, entry submissions, weekly review confirmations, recall displays, and reflection reports. Events contain only classification values such as tone, character-count range, number of items picked, recall type, report reason, and the type of the reported response (standard reflection, safety response, fallback response), and do not contain original sentences, drafts in progress, or reflection conversations

Google Analytics SDK (Google Analytics for Firebase; operates only if you consent; off by default)

Google’s global infrastructure (including the United States)

Device and app information

App instance ID, Firebase Installation ID (Firebase Installations ID), iOS Identifier for Vendors (IDFV, iOS only), device model, OS version, app version, language and region settings, session information, build type

Google Analytics SDK

Google’s global infrastructure (including the United States)

Approximate location

City-level location estimated from the IP address. The IP address is used only to estimate location at the time of collection and is discarded before it is logged

Google Analytics SDK

Google’s global infrastructure (including the United States)

Usage statistics (PostHog)

Usage events such as screen views and feature use, device and app information, identifier assigned by PostHog, approximate location estimated from the IP address. Does not contain original sentences, drafts in progress, or reflection conversations

PostHog SDK (operates only if you consent; off by default)

PostHog Cloud (United States)

Crash reports (Firebase Crashlytics)

When the App terminates abnormally or an unhandled error occurs: the error type, message, and stack trace, time of occurrence, device model, OS version, app version, and Crashlytics installation identifier. The App does not set an account identifier in reports, and reports do not contain original sentences or reflection conversations. However, if an error message includes a server document path, it may contain an account identifier

Firebase Crashlytics SDK (always active in release builds; no setting in the App to turn it off)

Google’s global infrastructure (including the United States)

Crash reports (Sentry)

When the App terminates abnormally or an unhandled error occurs: the error type, message, and stack trace, time of occurrence, device model, OS version, app version, build environment value (dev or prod), and event identifier assigned by Sentry. The App does not set an account identifier in reports, is configured not to store the IP address as a user attribute, and does not send screenshots or the App’s log output, and reports do not contain original sentences or reflection conversations. However, if an error message includes a server document path, it may contain an account identifier

Sentry SDK (always active in release builds; no setting in the App to turn it off)

Sentry (United States)

App integrity checks

A token confirming that a request comes from the genuine App (for security verification). The device attestation data used to issue the token is passed to Apple (App Attest, DeviceCheck) or Google Play Integrity, and App Check does not retain it

Firebase App Check

Google’s global infrastructure (including the United States)

Region code and app language

The country code from your device’s region setting (e.g., KR, JP, US) and the app language. The country code is used to select the support hotlines for safety responses, and the app language is used to determine the response language

Sent by the App with each reflection request

Not stored separately (for requests converted to a safety response, the region is recorded in server operation logs)

Call count records

Per-account counter of reflection requests, counter of monthly comment generations

Server

Cloud Firestore (United States)

Safety response records

Time of the most recent safety response and identifier of the entry to which that response is attached (does not include the crisis-signal screening result itself)

Recorded when the server sends a safety response (Section 7.3)

Cloud Firestore (United States) account document

Server operation logs

Account identifier (UID), function name, reason and response type of received reflection reports, crisis-signal screening information (keyword rule level, AI classification result, classification method, region — Section 7.3). Logs written by the Company do not record original sentences or response text. However, in exceptional errors, such as malformed requests, the request content may remain in the server framework’s error logs (Section 11.2)

Cloud Functions logs (Cloud Logging)

Google’s global infrastructure (including the United States)

Server request records

Request IP address, the App’s request information (User-Agent), time of request

Google Cloud default request logs (Cloud Logging)

Google’s global infrastructure (including the United States)

Sign-in records

IP address of sign-in requests. Recorded by Firebase Authentication and retained for several weeks (as published by Firebase)

Firebase Authentication

Firebase Authentication (United States)

All items in the table above whose collection method is the Google Analytics SDK or the PostHog SDK are collected only if you have consented to usage statistics. On the consent screen shown after sign-in and age confirmation, the Company informs you of the items collected, the purposes of use, the retention period, matters concerning cross-border transfer (the recipients, the destination countries, etc.), and your right to refuse consent, and obtains this consent separately from AI processing consent. The default is off, and you can use all features even if you do not consent. You can withdraw or give consent again at any time in the top bar under “Settings” > “Privacy” > “Send usage statistics”, and while you have not consented or have withdrawn consent, usage statistics are not sent to Google Analytics or PostHog. Usage statistics already sent before withdrawal are subject to the retention periods in Section 9.

The Company does not send original sentences, drafts in progress, reflection conversations, or crisis-signal screening results to Google Analytics or PostHog as analytics events. However, if you report a safety response, the response type (safety response) is included in the report event. Because the Company does not set an account identifier in Google Analytics, Google Analytics usage statistics are collected per app instance, not per account.

3.4 Information Collected from Third Parties (Sign-In Providers)

Source

Items collected

Google Sign-In

Account identifiers (Firebase UID, Google user ID), email address, display name and profile photo URL (to the extent provided by Google)

Sign in with Apple

Account identifiers (Firebase UID, Apple user ID), email address (Apple’s private relay address if you choose “Hide My Email”). Apple may provide your name to the App, but the Company does not store it on its servers (Firebase Authentication)

The above information is kept in Firebase Authentication. The Company does not create a separate profile document. Email/password, anonymous, and other social sign-in methods are not offered.

3.5 Information the Company Does Not Collect

  • Unique identification information, such as resident registration numbers and alien registration numbers
  • Precise location information (GPS), contacts, photos, and camera or microphone input
  • Payment information (the Service currently has no payments)
  • Advertising identifiers: the iOS Identifier for Advertisers (IDFA) and the Android Advertising ID (AAID) are not collected for usage statistics or advertising purposes. On iOS, the Identifier for Vendors (IDFV), which is not an advertising identifier, may be processed for analytics purposes (Section 3.3)
  • Push notification tokens (the App does not send push notifications)
  • Device accelerometer values — read only on the device for the tilt effect on the home screen, and neither stored nor transmitted
  • Transmission of your email, name, or account identifier to AI processing providers — reflection requests do not contain such account information. However, names and other details that you yourself write in a sentence are sent as they are, together with the sentence

The Company does not ask for sensitive information such as health or mental health information. However, if you have consented to AI reflections, the Service automatically screens sentences for signals of self-harm or suicide risk in order to provide safety responses. The handling of sensitive content you leave in sentences, and crisis-signal screening, are governed by Section 7.

3.6 Information of Persons Under the Minimum Age

The Service is available only to persons aged 16 or older, and right after Google or Apple sign-in, the App asks you to confirm that you are 16 or older. If you answer that you are under 16, you cannot use the Service and are signed out, and that answer is not recorded. Section 18 explains what information remains even after you answer that you are under 16, and how it is handled.

3.7 Website and Launch Notification Sign-Ups

Category

Items processed

Purpose

Retention basis

Launch notification sign-ups

Email address you enter, Website language, referral source. When you sign up, the request’s IP address, browser information (User-Agent), and referring page may be stored together with the sign-up record

Sending StarWrite launch news and managing sign-ups

Deleted without delay when you request deletion, or when it is no longer needed because the purpose of the launch announcement has been achieved

Website usage statistics

Pages visited, referral source, browser and device information (Cloudflare Web Analytics; uses no cookies and does not identify individuals)

Understanding Website performance and usage

Subject to Cloudflare’s service policies

Launch notification sign-up information is stored in the database service of Cloudflare, Inc. (Cloudflare D1, Western United States). To have your sign-up information deleted, please send a request to contact@passionfactory.ai. If the Company intends to use launch notification sign-up email addresses for marketing other than StarWrite launch announcements, it will follow separate notice or consent procedures in accordance with relevant laws.

The Website is provided through the services of Cloudflare, Inc., and visit statistics are aggregated only with Cloudflare Web Analytics. The Website does not use other analytics tools such as Google Analytics or PostHog. Cookies used by the Website are described in Section 15.3. The Company does not use the original text of your StarWrite entries or AI conversations for web analytics.

4. How Personal Information Is Collected

  • Received from the sign-in provider during Google or Apple sign-in
  • Entries, weekly review selections, age confirmation, AI processing consent, settings, and reflection reports that you enter directly in the App
  • Information generated automatically while you use the App (Firebase App Check, Firebase Crashlytics and Sentry crash reports, server operation logs, call count records, safety response records, and, if you have consented to usage statistics, Google Analytics and PostHog)
  • The device’s country code and the app language, which the App sends with each reflection request
  • Inquiries you send to the Company’s email
  • Information you provide directly through collection screens on the Website, such as launch notification sign-ups
  • Website performance and visit statistics collected through the analytics tool applied to the Website (Cloudflare Web Analytics)

The operation of automatic collection tools and how to refuse them are explained in Section 15.2.

5. Purposes of Processing Personal Information

5.1 Processing by Purpose

Purpose of processing

Details

Parties involved

Account creation and sign-in

Google and Apple sign-in, account identification, age confirmation

Google LLC (Firebase Authentication), Apple Inc. (sign-in provider; not a processor)

Storing and displaying entries

Storing sentences, displaying stars and galaxies, keeping entries across multiple devices, weekly reviews

Google LLC (Cloud Firestore)

AI reflections (if you consent)

Generating reflection responses to sentences; checking whether generated responses followed the conversation rules (no questions, no advice, etc.). Where recall is provided, quoting sentences from entries picked in past weekly reviews (recall is not provided as of the Reference Date; Section 8.7)

Google LLC (Cloud Functions), Vercel Inc. (Appendix A.2), Claude model provider (Appendix A.3), TypeSafe AI (response-rule check, Appendix A.4)

Crisis-signal safety responses (if you consent to AI reflections)

Screening today’s sentence for signals of self-harm or suicide risk and, if a signal is present, providing a safety response containing regional support hotlines instead of an AI response; for 30 days thereafter, suspending recall and replacing the monthly comment (Section 7.3)

Google LLC (Cloud Functions, Cloud Firestore), Vercel Inc. (Appendix A.2), Claude model provider (Appendix A.3)

Monthly comments (if you consent to AI reflections)

When a constellation is completed, generating and storing a one-line comment from the entries of that period (Section 8.10)

Google LLC (Cloud Functions, Cloud Firestore), Vercel Inc. (Appendix A.2), Claude model provider (Appendix A.3)

Handling reflection reports

Operator review of reported responses; improving response quality and safety

Google LLC (Cloud Firestore)

Service stability and security

Issuing app integrity check tokens (not currently enforced by the server), limiting excessive calls, diagnosing errors through crash reports, preventing abuse

Google LLC (Firebase App Check, Firebase Crashlytics, Cloud Functions), Functional Software, Inc. (Sentry)

Service improvement (usage statistics, if you consent)

Analyzing metrics such as entry usage rates and weekly review participation rates

Google LLC (Google Analytics), PostHog Inc. (PostHog)

Responding to inquiries and handling requests to exercise rights

Answering email inquiries; handling requests for access, correction, and deletion

Handled by the Company itself

Launch announcements

Sending StarWrite launch announcements to users who signed up for launch notifications on the Website

Cloudflare, Inc. (Cloudflare D1)

Website provision and usage statistics

Providing the Website; understanding Website performance and usage

Cloudflare, Inc.

Legal compliance

Fulfilling legal obligations, responding to lawful requests, handling disputes

-

5.2 Purposes for Which the Company Does Not Use Personal Information

Purpose not pursued

Explanation

Sale

The Company does not sell your personal information or entries. Transfers resulting from a merger, business transfer, or the like are governed by Section 16.2

Advertising

The Company currently does not display advertising and does not share personal information with third parties for advertising. If the Company introduces advertising in the future, it will update this Policy and obtain any consent required by law before doing so

AI model training

The Company does not use what you enter for training, retraining, or additional training (fine-tuning) of AI models. The Company also applies settings that prohibit training use and require zero data retention (ZDR) to AI processing providers (Section 8.5)

Profiling

The Company does not create profiles that infer your tendencies or emotional state. Crisis-signal screening (Section 7.3) does not infer your characteristics but is a safety measure concerning a single sentence, and its result is used only to provide a safety response and, for 30 days thereafter, to refrain from recall and to replace the monthly comment with a predefined message. The monthly comment (Section 8.10) is a sentence that puts into words the tone statistics of that month’s entries and does not infer tendencies or emotional state

Automated decisions with legal effect

The Company does not make automated decisions that produce legal effects or similarly significant effects on you (Section 8.8)

6. Legal Bases for Processing Personal Information

6.1 Users in the Republic of Korea (Under PIPA)

Processing activity

Legal basis

Account creation and sign-in; storing and displaying entries

Entering into and performing a contract (Article 15(1)(iv) of PIPA)

AI reflections, crisis-signal screening and safety responses, monthly comments

Consent of the data subject (Article 15(1)(i); the AI processing consent screen right after sign-in and age confirmation)

Handling reflection reports

Performing a contract and taking measures at the data subject’s request (Article 15(1)(iv))

Service stability and security (including crash reports)

Legitimate interests of the personal information controller (Article 15(1)(vi))

Service improvement (usage statistics)

Consent of the data subject (Article 15(1)(i); the in-app usage statistics consent screen)

Responding to inquiries and handling requests to exercise rights

Measures taken at the data subject’s request (Article 15(1)(iv)); fulfilling legal obligations (Article 15(1)(ii))

Overseas storage and processing essential to the Service (Google Firebase, excluding usage statistics)

Outsourcing of processing and storage for entering into and performing a contract, disclosed in this Policy (Article 28-8(1)(iii))

Cross-border transfer for usage statistics (Google Analytics, PostHog)

Consent of the data subject (Article 28-8(1)(i); matters concerning the cross-border transfer are also disclosed on the in-app usage statistics consent screen)

Cross-border transfer for AI processing (Vercel, Claude model provider, TypeSafe AI)

Consent of the data subject (Article 28-8(1)(i); the in-app AI processing consent screen)

Website launch notification sign-ups

Consent of the data subject (Article 15(1)(i); the sign-up screen)

Website usage statistics

Cloudflare Web Analytics: legitimate interests of the personal information controller (Article 15(1)(vi); visit statistics aggregated without cookies)

6.2 Users in Japan (Under APPI)

The Company processes personal information within the scope of the purposes of use published in this Policy, and obtains your prior consent if use beyond those purposes is required. Information on provision to third parties in foreign countries (such as processors located in the United States; usage statistics in Google Analytics, app integrity checks, crash reports, server operation logs, and AI processing may also be processed outside the United States) is provided in Sections 13 and 17.2.

6.3 Users in California, United States (Under the CCPA)

The Company collects and uses personal information only for the purposes in Section 5.1, and does not go beyond what is reasonably necessary and proportionate to achieve those purposes. The categories collected and your rights are explained in Section 17.3.

7. Entries Containing Sensitive Content

7.1 The Company Does Not Ask for Sensitive Information

The Company does not ask for sensitive information under Article 23 of PIPA (health, ideology or beliefs, sex life, etc.) for use of the Service, and all features of the Service work properly without sensitive information.

7.2 What Entries May Contain

Because StarWrite is a review app for looking back on your day, the sentences you leave may, at your own discretion, contain content about your emotional state, health, or mental health. Whether to leave such content is entirely your voluntary choice. Such sentences are stored and protected in the same way as other entries (Section 11), and are sent along the AI processing path only if you have consented to AI reflections (Section 8).

7.3 Crisis-Signal Screening and Safety Responses

When a user who has consented to AI reflections leaves a sentence, the Service automatically screens today’s sentence for risk signals suggesting self-harm or suicide. If a risk signal is detected, the Service shows a predefined safety response and regional crisis support hotlines (Republic of Korea, United States, Japan, etc.) instead of an AI-generated response. This screening is not profiling that infers your characteristics or tendencies but a safety measure concerning that single sentence. There is no setting to turn off crisis-signal screening alone; if you decline or withdraw consent to AI reflections, your sentences are not passed to the screening process, so neither screening nor safety responses are provided.

Item

Details

Screening method

Keyword rules and an AI classifier are used together. For AI classification, today’s sentence is sent along the AI processing path (through Vercel AI Gateway to the Claude model provider) one more time, separately from reflection generation (Section 8.2). If a strong crisis keyword is present, screening is completed first; for other sentences, screening and reflection generation proceed simultaneously, so a sentence screened as containing a risk signal may also be sent to the reflection-generation model, in which case the generated response is discarded and a safety response is sent

Selection of support hotlines

Selected based on the country code from the device region setting sent by the App; if there is no country code, the region assigned to the account is used, and if there is none, U.S. hotlines are shown. The hotlines shown may differ from where you currently are

Handling of the entry

Whatever you write, your sentence is not rejected or deleted, and it still becomes a star

Information kept with the account

The crisis-signal screening result itself (keyword rule level, AI classification result) is not stored in account data. However, when a safety response is sent, the time of the most recent safety response and the identifier of the entry to which it is attached are recorded in the account document, and the response is stored in the reflection conversation as a safety response (response type “safety response”). This information is deleted together with your account when you delete it

The following 30 days

For 30 days from the time of the most recent safety response, no recall that revisits past sentences is made (Section 8.7), and the monthly comment for any constellation completed during that period is replaced with a predefined message instead of being generated by AI (Section 8.10). Conversations that included a safety response are excluded from the context of subsequent AI requests (Section 8.3)

Server operation logs

For requests that triggered a keyword rule or that AI classification judged to indicate a crisis, the account identifier, keyword rule level, AI classification result, and classification method are logged, and for requests converted to a safety response, the region used to select the support hotlines is also logged. The original sentence and the matched keywords themselves are not logged, and logs are automatically deleted after 30 days (Section 9.1)

Reports and usage statistics

If you report a safety response, the response type (safety response) is included in the report document and the report event (Section 8.9). Otherwise, crisis-signal screening results are not sent as usage statistics

Notifications

The Service does not notify the Company, guardians, or third parties on the basis of your entries or screening results, and does not request emergency rescue

Monitoring

The Company and its officers and employees do not monitor your entries in real time, and no person checks screening results in real time

7.4 When You Need Help

  • StarWrite is not a medical or counseling service. Reflections and safety responses do not replace professional help.
  • Because crisis-signal screening is an automated tool, it may miss signals or judge a sentence that does not indicate a crisis to be a crisis.
  • If you feel you are in danger right now, do not rely on StarWrite; please seek help immediately from emergency services, medical institutions, a trusted guardian, or professional support organizations where you are.

8. AI Reflections and Personal Information Processing

8.1 Overview of the AI Reflection Feature

When you leave a sentence, if you have consented to AI processing, the Universe (the narrator within the Service) shows you a response that briefly reflects that sentence back to you, that is, a reflection. Reflections are an optional feature, and even if you do not consent, you can continue to use other features such as entries and star creation.

8.2 Processing Path

Step

Party

Location

Role

1

Your App

Your device

Sends the reflection request. The App contains no API keys for AI processing providers

2

StarWrite server (Google Cloud Functions)

United States

Relays AI requests after authentication, consent verification, and call-count limiting. Crisis-signal screening proceeds alongside reflection generation, and if a risk signal is detected, the generated response is discarded and a safety response is sent (Section 7.3)

3

Vercel Inc. (AI Gateway)

United States and other regions operated by Vercel

Relays AI requests and applies the settings that prohibit training use and require zero data retention. Does not store prompts or responses

4

Claude model provider: Anthropic, PBC, or a Claude hosting provider to which the gateway routes the request (such as Claude hosted by a cloud provider)

Regions operated by the provider, including the United States (not fixed to the United States)

Generates reflection responses, classifies crisis signals, and generates monthly comments. Does not store the sentences sent

5

TypeSafe AI (response-rule check model, via AI Gateway)

Appendix A.4

Checks whether the generated reflection response followed the Company’s conversation rules (no questions, no advice, etc.). Does not store the sentences and responses sent

The models currently used are from Anthropic’s Claude family. Reflection requests are passed to the Claude model through Vercel AI Gateway (the United States and other regions operated by Vercel). Because the Company does not fix the model provider or inference region in AI Gateway, the gateway may, depending on availability and other factors, route requests to another hosting provider that offers the same Claude model (e.g., Claude hosted by a cloud provider), and the processing region is not fixed to the United States. Whichever path processes a request, the Company applies zero data retention (ZDR) and training-prohibition settings to each request, and requests are not passed to any provider to which zero data retention does not apply (Section 8.5). If the configuration of AI processing providers, model providers, or processing regions changes, the Company will update this Policy and notify you, and will obtain consent for any change that requires consent under law.

8.3 Items Sent

Item

Description

Today’s sentence

The original text of the sentence for which you receive a reflection

Previous reflection conversation

The message and response history (up to 20 turns) of the single most recent conversation stored on the server, to maintain context. Conversations that included a safety response are excluded

Recall quotations

(Where recall is provided) the original text of sentences from entries you picked in weekly reviews. Recall is not provided as of the Reference Date

Crisis-signal classification request

Today’s sentence. Sent one more time, separately from reflection generation, to screen for signals of self-harm or suicide risk (Section 7.3)

Entries of a completed constellation

When a constellation is completed, the original text of the sentences of entries left during that period (up to 28), their dates and tones, and the constellation name are sent once to create the monthly comment (Section 8.10)

Response-rule check request

The text of the generated reflection response, together with today’s sentence, your messages in the most recent conversation (including quoted sentences where recall is provided), and the response language, which serve as the basis for the check. Sent to TypeSafe AI to check whether the response followed the Company’s conversation rules (Appendix A.4)

Items not sent

Account information such as email, name, account identifier, and IP address is not sent to AI processing providers. However, names and other details that you yourself write in a sentence are sent as they are, together with the sentence

With each reflection request, the App also sends the country code from the device’s region setting and the app language. The country code is used to select support hotlines for safety responses and the app language is used to determine the response language; the country code is not passed to AI processing providers. Whether you have given AI processing consent is not determined from a value sent by the App; the server checks the account’s consent record directly (Section 8.4). The IP address and User-Agent of requests received by the server are recorded in Google Cloud request logs (Section 3.3).

8.4 Consent and Withdrawal

Item

Details

When consent is requested

On the “May we reflect on the sentences you write?” screen shown right after Google or Apple sign-in and age confirmation, you choose “Yes, please reflect” or “Not right now”. If you move past this screen without deciding, the same notice is shown again on the home screen

What the consent screen discloses

Before obtaining AI processing consent, the Company informs you, on the consent screen or in linked detailed information, of the AI processing providers, the countries and regions where processing may take place, the items sent (today’s sentence, previous reflection conversation, recall quotations), the purposes of processing, the retention and use criteria, the non-use of data for AI training, and how to refuse or withdraw consent and the effects of doing so.

When the App asks again

The App asks only when there is no consent record on the server. A decision to consent or decline carries over to new devices, and if you declined, you are not asked again

Consent record

Your decision is recorded only on the server (Cloud Firestore, a consent document under your account), and only a cache for offline use is kept on your device. The server record contains only the decision, the time, and the version of the consent notice, and does not contain original sentences

Server verification

With each reflection request, the server checks the consent record on the server, and rejects the request if there is no consent record, if consent has been declined, or if the record cannot be verified

How to withdraw

You can withdraw at any time by turning off “Send my sentences to AI for reflections” in the top bar under “Settings” > “Privacy”. Turning it back on restores your consent

Effect of withdrawal

Withdrawal applies from your next sentence. Requests already sent and processed, and responses already generated, cannot be undone. However, under the zero data retention settings, the sentences sent are not stored by AI processing providers (Section 8.5). Because the server checks the consent record on the server with each reflection request, withdrawal takes effect once it is recorded on the server, which usually happens within a few seconds. If updating the server record is delayed due to network problems, reflections may continue on devices of the same account, including this device, until the record is updated

While declined or withdrawn

AI is not called, and the fixed message “I’ll keep this. Even without a reflection, your star still shines.” is shown

Retention of the consent record

Upon withdrawal, the consent record on the server is not deleted but overwritten with “declined” (for cases where the server update is delayed, see “Effect of withdrawal” above), and it is deleted together with your account when you delete it

8.5 Prohibition of AI Training Use and Zero Data Retention

Item

Details

No training use by the Company

The Company does not use the sentences you enter or reflection conversations for training, retraining, or additional training (fine-tuning) of AI models, or for building evaluation datasets

Training prohibition specified to providers

The server specifies AI Gateway’s option prohibiting training on prompts for every AI request. Vercel states that it does not use prompts for training

Zero data retention on the provider side

The Company applies zero data retention (ZDR) settings to all AI requests, including reflection generation, crisis-signal classification, monthly comment generation, and response-rule checks, so that the sentences and conversations sent are used only for processing and are not stored by Vercel AI Gateway, the Claude model provider, or TypeSafe AI. The Company does not use any model or path to which zero data retention does not apply

Operational metadata

Operational metadata that does not contain sentence or conversation content (status, model, token usage, cost, duration, etc.) may be recorded separately in accordance with the AI processing providers’ operational policies (Section 9.1)

8.6 Temporary Retention on the Server

So that you can resume receiving a reflection response even if the connection drops, the server briefly keeps the text of the reflection response being generated in Cloud Firestore. Your sentence is not subject to this temporary retention. Temporary retention documents are kept only for the short period needed to recover the connection and then deleted, and are deleted immediately from active systems upon account deletion. The App cannot access these documents directly.

Separately from the temporary retention above, the reflection conversation history itself is stored on the server (Cloud Firestore, per entry under your account), is used as context for the next reflection (the single most recent conversation, up to 20 turns — Section 8.3), and remains available to view even if you sign in on another device. Reflection conversations are deleted together with your account when you delete it. If you sign out, the cache on this device is deleted, but the reflection conversations on the server remain.

8.7 Recall

Recall is a feature that adds to a reflection a sentence that gently revisits a past entry, and it takes two forms.

  1. Tone count sentence: A single sentence telling you how many times the same tone appeared among your recent entries. The server only counts tones and builds the sentence from fixed wording, and it is not sent to AI processing providers.
  2. Quotation of picked entries: The server reads the original text of the sentences of entries picked in weekly reviews and includes them in the AI request as reference quotations, and the AI quotes part of them in the reflection response.

Recall is not provided as of the Reference Date, and the Company will update this Policy before it begins providing recall. When provided, it may apply only to some users depending on server settings. The table below sets out the processing criteria that apply when recall is provided.

Item

Details

What may be quoted

Up to the 8 most recent entries you picked in weekly reviews. Entries with the tone “heavy” are excluded

Frequency

Each picked entry is quoted only once

Exclusions

No recall is made if there is no AI processing consent, or if 30 days have not passed since the time of the most recent safety response (Section 7.3)

Retention of sentences

The server record of weekly review selections contains only which entries were picked and whether recall was used; the sentences to be quoted are read again from the entries each time a request is made

8.8 Automated Decision-Making

Reflections and crisis-signal screening are not decisions that produce legal effects or similarly significant effects, and they are not used for other determinations such as restricting account use. Crisis-signal screening results are used only to provide safety responses and for the 30-day handling described in Section 7.3. Reflections are responses for reference only, and their factual accuracy is not guaranteed.

8.9 Reflection Reports

If a reflection response bothers you, you can report it directly in the App by long-pressing the response bubble.

Item

Details

Items stored

Text of the reported response, report reason, response type (standard reflection, safety response, fallback response), identifier of the relevant entry, time of the response, time of the report

Items not stored

The original text of your entry is not separately sent or stored. However, if you report a response containing recall, part of the sentence quoted in the response text may be stored with it

Access

You yourself cannot read or modify report documents in the App, and only authorized operators review them. If necessary for the review, an operator may open the entry (sentence) to which the reported response is attached

Retention

Retained while your account exists, and deleted together with your account when you delete it

Usage statistics

Only the report reason and the response type (including safety responses) are sent as usage statistics; the response text and the entry identifier are not included

8.10 Monthly Comments for Completed Constellations

When a constellation is completed, the server sends the entries from that constellation’s period once to the AI processing provider to create a one-line comment that puts into words the tone statistics of that month’s entries, and stores the comment in the constellation record. The same processing path as for reflections (Section 8.2) and the same training-prohibition and zero data retention conditions (Section 8.5) apply. The monthly comment describes that month’s entry statistics and is not profiling that infers your tendencies or emotional state.

Item

Details

Items sent

The original text of the sentences of entries left during that constellation’s period (up to 28), the date and tone of each entry, and the constellation name. Entries that received a safety response are not sent

Frequency and limits

Created only once per constellation; requests are limited to 10 per minute and 30 per day per account

When AI is not used

If there is no AI processing consent, if you have not consented under the current version of the consent notice that describes monthly comments, if 30 days have not passed since the time of the most recent safety response (Section 7.3), if fewer than 7 entries can be sent, or if the app language is not Korean, sentences are not sent to the AI processing provider and a predefined message is stored instead

Storage

The comment is stored in the constellation record in Cloud Firestore (United States), and deleted together with your account when you delete it

Shared card

You can export a card image containing a completed constellation’s name, period, star shape, and monthly comment outside the App using your device’s share function. The card does not contain original sentences, and the Company does not receive the exported image. Once exported, managing the image is your responsibility

Logs

Server operation logs do not record original sentences or the comment text

9. Retention and Use Periods

9.1 Retention Periods by Item

Data category

Retention period

Account information (Firebase Authentication)

While the account exists, until the account is deleted (after deletion, removed from Firebase Authentication’s operational and backup systems within 180 days). The IP address of sign-in requests is retained for several weeks (as published by Firebase)

Entries (sentence, time, date, tone)

While the account exists. Deleted immediately upon account deletion

Weekly reviews and segment reviews (identifiers of picked entries, times, number of times you postponed the invitation)

While the account exists. Deleted immediately upon account deletion

Reflection conversations (message and response history, response type)

While the account exists. Deleted immediately upon account deletion

Constellations (shape, name, monthly comment, position information, time constellation guidance was shown, number of times declined)

While the account exists. Deleted immediately upon account deletion

Account document (age confirmation time, day boundary time, time of the most recent safety response and the identifier of that entry)

While the account exists. Deleted immediately upon account deletion

Submission processing records (identifiers of the representative entry for each date and of submission requests, and processing status)

While the account exists. Deleted immediately upon account deletion

AI processing consent record (server)

While the account exists (overwritten with “declined” upon withdrawal). Deleted immediately upon account deletion

Reflection reports

While the account exists. Deleted immediately upon account deletion

Temporary retention of reflection responses

Deleted after short-term temporary retention for connection recovery. Deleted immediately from active systems upon account deletion

Call count records

While the account exists. Deleted immediately upon account deletion

Sentences sent to AI processing providers

Not stored. The Company applies zero data retention (ZDR) and training-prohibition settings to all AI requests, and sentences and conversations are processed only while a response is being generated. Operational metadata that does not contain sentence or conversation content (status, model, token usage, cost, duration, etc.) may be retained separately in accordance with the AI processing providers’ operational policies

Server operation logs

30 days (the retention period of the Cloud Logging default bucket). Automatically deleted after this period

Server request records (IP address, User-Agent)

30 days (the retention period of the Cloud Logging default bucket). Automatically deleted after this period

Usage statistics (Google Analytics, PostHog — if you consent)

Subject to the analytics data retention period set by the Company and the service policies of each analytics tool

Crash reports (Firebase Crashlytics)

90 days (Google policy). Because no account identifier is set, they are not deleted together with the account upon account deletion

Crash reports (Sentry)

90 days (Sentry default retention policy). Because no account identifier is set, they are not deleted together with the account upon account deletion

App integrity check tokens (Firebase App Check)

Tokens are valid for up to 7 days, and tokens used to prevent replay are retained for up to 30 days. Device attestation data is not retained by App Check

Server backup and recovery copies

Automatically destroyed after the retention period set in the Company’s backup and recovery policy

Email inquiry records

Deleted without delay once the purpose of handling the inquiry has been achieved. However, retained for the necessary period if there is a legal retention obligation or a need to respond to a dispute

Website launch notification sign-ups

Deleted without delay once the purpose of the launch announcement has been achieved or upon a deletion request. If used for a separate marketing purpose in the future, notice and consent procedures under relevant laws will be followed

Information subject to legal retention obligations

Information that relevant laws require to be retained (e.g., payment and transaction records) is retained to the minimum extent necessary, separately from other information, for the period prescribed by the relevant law and then destroyed. As of the Reference Date, the Service has no payments, so no payment records are retained

Local data on your device

Until you sign out, delete your account, or delete the App. When you sign out or delete your account, the device cache of server data (including age confirmation, consent decision, and day boundary) is also deleted. However, device settings such as the usage statistics setting and the app language remain on the device until the App is deleted (Section 15.1)

9.2 Processing After Account Deletion

Step

Processing

Disconnection from sign-in providers

For Apple sign-in, the token is revoked; for Google sign-in, the App’s connection to your Google account is removed

Deletion of server data

The account document (including age confirmation, day boundary, constellation guidance, and safety response records), entries, reflection conversations, AI processing consent record, weekly reviews and segment reviews, constellations (including names and monthly comments), submission processing records, reflection reports, temporarily retained reflection responses, and call count records are deleted immediately, with no grace period

Account deletion

The account in Firebase Authentication is deleted

Irreversible

Deleted accounts and server data cannot be restored. If you sign in again with the same Google or Apple account, you start with a new account

Deletion of device data

The account data on this device (drafts in progress, submissions whose transmission was not completed, and the offline cache of server data, including reflection conversations, weekly reviews, and consent records) is deleted, and you are signed out

9.3 Information Not Deleted Immediately upon Account Deletion

Subject

Handling

Server backup and recovery copies

Not deleted directly by the account deletion function; they disappear once the backup retention period expires. The Company does not reuse such copies to provide the Service in the ordinary course. If data is restored from a backup, accounts and entries deleted before the restoration are deleted again immediately after the restoration

Sign-in account information (Firebase Authentication)

Removed from Firebase Authentication’s operational and backup systems within 180 days after deletion. The IP address of sign-in requests is retained for several weeks (as published by Firebase)

Usage statistics

Event data already sent to analytics tools (Google Analytics, PostHog) is subject to the retention settings of each analytics tool. No account identifier is set in Google Analytics

Server operation logs and request records

Automatically deleted once the 30-day log retention period has passed (Section 9.1)

Information subject to legal retention obligations

When you delete your account, your information such as entries is deleted, but information that relevant laws require to be retained (e.g., payment and transaction records) is retained separately to the minimum extent necessary for the period prescribed by the relevant law and then destroyed (Section 9.1)

Account deletion marker (tombstone)

Account deletion is available only while online. To prevent late-arriving writes from restoring data immediately after deletion, the server keeps in Cloud Firestore, for up to 1 hour, a deletion marker containing only the account identifier, which blocks storage for that account

Crash reports

Crash reports already sent to Firebase Crashlytics and Sentry are not linked to the account and are deleted after their retention period (90 days) (Section 9.1)

Device settings on this device

Device settings such as the usage statistics setting and the app language remain on the device. The device cache of server data, such as age confirmation, AI processing consent, and day boundary, is deleted together with the account. If you delete the App, they are removed along with the data in the App’s dedicated storage (for the iOS Keychain exception, see Section 15.1)

10. Procedures and Methods for Destroying Personal Information

The Company destroys personal information without delay once its retention period has ended or the purpose of processing has been achieved. Electronic data on the server is deleted from the operational database so that it cannot be restored, and backup copies are automatically destroyed once their retention period has expired (see Section 9.3, including the reapplication of deletions upon restoration from backup). Inquiry records received by email are deleted once their retention period has passed. Data stored on your device is deleted by the App when you sign out or delete your account, or is deleted by the operating system together with the App’s dedicated storage when you delete the App (for the iOS Keychain exception, see Section 15.1).

11. Measures to Ensure the Security of Personal Information

11.1 Administrative Safeguards

  • Designation and operation of a Chief Privacy Officer
  • Officers and employees are prohibited from viewing entries in the ordinary course. Authorized personnel access entries only to the minimum extent necessary and only in very limited cases, such as at your request (including reviewing reports and handling requests to exercise your rights), responding to outages or security incidents, or fulfilling legal obligations (this is a control based on operational policy; entries are not end-to-end encrypted)
  • Centralized management of, and access control over, secrets such as API keys for AI processing providers

11.2 Technical Safeguards

Category

Measures

Encryption in transit

HTTPS (TLS) is applied to all communications between the App and the server and between the server and the AI processing path

Protection of stored data

Google Cloud’s default encryption of stored data is used

Access control

Database security rules deny all access not explicitly permitted, and from the App you can access only your own data. Reflection reports, submission processing records, and the server’s temporary retention documents cannot be accessed even from your App. Operator access is limited to restricted purposes (Appendix B)

App integrity checks

The App obtains app integrity check tokens through Firebase App Check and attaches them to requests, but as of the Reference Date, the server does not enforce these tokens. If enforcement is introduced, this Policy will be updated

API key isolation

API keys for AI processing providers are kept only on the server and are not included in the App

Server-side consent verification

The server rejects requests for which AI processing consent cannot be verified

Log minimization

Server logs written by the Company do not record original sentences or response text. However, in exceptional errors, such as malformed requests, the request content may remain in the server framework’s error logs, and such logs are deleted once the log retention period (Section 9.1) has passed

Analytics minimization

Usage statistics are collected only if you consent, and events contain only classification values, not original sentences, drafts in progress, reflection conversations, or crisis-signal screening results. No account identifier is set in Google Analytics

Call limits

The number of reflection requests per account is limited to prevent abuse

11.3 Physical Safeguards

As of the Reference Date, the Google Cloud data centers used by the Company (servers and databases in the United States; usage statistics, app integrity checks, and server operation logs on Google’s global infrastructure, including the United States) apply industry-standard physical security measures (access control, continuous monitoring, etc.). Physical protection for the AI processing path (Vercel, the Claude model provider, TypeSafe AI), PostHog, and Cloudflare is subject to each provider’s security measures.

11.4 Notification of Personal Information Breaches

Applicable law

Notification

PIPA (Republic of Korea)

Notify data subjects within 72 hours of becoming aware of a leak or similar incident, and report to the Personal Information Protection Commission or the Korea Internet & Security Agency in accordance with the criteria prescribed by law

APPI (Japan)

Report to the Personal Information Protection Commission of Japan and notify the individuals concerned in accordance with the criteria prescribed by law

California and other U.S. states

Notify at the time and in the manner prescribed by the laws of the relevant state

12. Outsourcing of Processing

12.1 Processors and Outsourced Tasks

The Company outsources the processing of personal information to the following processors in order to provide the Service. Details are set out in Appendix A, and any changes will be announced through this Policy or an in-app notice.

Processor

Outsourced tasks

Processing location

Google LLC (Firebase Authentication)

Sign-in authentication, account management

United States

Google LLC (Cloud Firestore)

Storing entries, reflection conversations, consent records, weekly reviews, constellations, account settings, and reports; temporary retention of reflection responses

United States

Google LLC (Cloud Functions, Cloud Logging)

Running server functions (relaying reflections, crisis-signal screening, monthly comment generation, account deletion, receiving reports); operation logs and request records

Cloud Functions: United States. Cloud Logging: Google’s global infrastructure (including the United States)

Google LLC (Google Analytics)

Aggregating usage statistics (only for users who have consented to usage statistics)

Google’s global infrastructure (including the United States)

PostHog Inc. (PostHog)

Aggregating usage statistics (only for users who have consented to usage statistics)

United States (PostHog Cloud)

Google LLC (Firebase App Check)

App integrity checks

Google’s global infrastructure (including the United States)

Google LLC (Firebase Crashlytics)

Collecting and analyzing app crash reports

Google’s global infrastructure (including the United States)

Functional Software, Inc. (Sentry)

Collecting and analyzing app crash reports

United States (Sentry)

Vercel Inc. (AI Gateway)

Relaying AI requests (only for users who have consented to AI processing)

United States and other regions operated by Vercel

Claude model provider: Anthropic, PBC, or a Claude hosting provider to which the gateway routes the request

Reflection generation, crisis-signal classification, monthly comment generation (only for users who have consented to AI processing)

Regions operated by the provider, including the United States (not fixed to the United States). Does not store the sentences sent

TypeSafe AI

Checking reflection responses for compliance with the conversation rules (only for users who have consented to AI processing)

Appendix A.4. Does not store the sentences and responses sent

Cloudflare, Inc.

Providing the Website (hosting and delivery), aggregating Website visit statistics (Cloudflare Web Analytics), storing launch notification sign-up information (Cloudflare D1)

Cloudflare’s global network (including the United States). Launch notification sign-up information: Western United States

With Google Sign-In and Sign in with Apple, authentication is performed by the sign-in provider you choose, and that provider’s privacy policy also applies (Appendix A.5).

The processors and processing locations above reflect the configuration as of the Reference Date and may change depending on future infrastructure configuration. If they change, the Company will update this Policy and notify you, and will follow the applicable procedures for any change that requires consent or notice under law.

12.2 Oversight of Processors

The Company oversees its processors so that they process personal information only within the scope of the outsourced tasks, and applies training-prohibition and zero data retention settings to each request sent to AI processing providers. Contractual oversight measures are governed by the data processing terms that apply between the Company and each processor. If the Company adds or changes processors, it will update this Policy and notify you.

12.3 Distinction Between Outsourcing of Processing and Provision to Third Parties

Outsourcing of processing to processors means that the Company entrusts processing to them in order to provide the Service; it is not a provision of personal information to processors for their own purposes. The Company does not provide personal information to third parties for marketing or advertising purposes.

13. Cross-Border Transfers of Personal Information

13.1 Overview of Cross-Border Transfers

As of the Reference Date, StarWrite’s servers and databases are located in the United States (this may change depending on future infrastructure configuration, and if it changes, this Policy will be updated). Accordingly, the personal information of users in the Republic of Korea and Japan is transferred to the United States as soon as they use the Service. App integrity checks, crash reports, and server operation logs and request records are processed on Google’s global infrastructure (including the United States). Crash reports are also sent to Sentry (United States). Usage statistics are transferred to Google Analytics (Google’s global infrastructure, including the United States) and PostHog (United States) only if you consent. Data for AI processing is transferred, only if you consent, through Vercel AI Gateway (the United States and other regions operated by Vercel) to the Claude model provider (Anthropic, PBC, or a Claude hosting provider to which the gateway routes the request) and TypeSafe AI, and the processing region is not fixed to the United States (Section 8.2). Transfers from the App fall into three categories, and the Website is provided through the global network of Cloudflare, Inc.

  • Transfers essential to the Service (Google, Sentry): Transfers for sign-in, entry storage, server functions (including server operation logs), app integrity checks, and crash reports (app integrity checks, crash reports, and server operation logs are processed on Google’s global infrastructure (including the United States), and Sentry crash reports in the United States). As outsourcing of processing and storage necessary to provide the Service, they are disclosed through this Policy and the Terms of Service.
  • Usage statistics transfers (Google Analytics, PostHog): Transfers for improving the Service. They take place only if you consent to usage statistics, and you can withdraw consent at any time in the settings (Sections 13.3 and 13.4).
  • AI processing transfers (Vercel, Claude model provider, TypeSafe AI): Transfers for AI reflections, crisis-signal classification, monthly comment generation, and response-rule checks. They take place only if you give separate consent on the AI processing consent screen right after sign-in and age confirmation.

13.2 Status of Cross-Border Transfers (Article 28-8 of PIPA)

Recipient

Destination country

Time and method of transfer

Items transferred

Purpose of transfer

Retention and use period

Google LLC — Firebase Authentication

United States

At sign-in / HTTPS

Account identifier, email address, display name and profile photo URL (Google sign-in only), IP address of sign-in requests

Sign-in authentication, account management

Until the account is deleted (after deletion, removed from operational and backup systems within 180 days). IP addresses are retained for several weeks (as published by Firebase)

Google LLC — Cloud Firestore

United States

As needed when entries, consents, and reports are made / HTTPS

Entries (sentence, time, date, tone), reflection conversations (message and response history), weekly reviews and segment reviews, constellations (name, monthly comment, position information), account document (age confirmation time, day boundary time, constellation guidance records, safety response records), submission processing records, AI processing consent record, reflection reports, temporarily retained reflection responses, call count records

Storing and displaying entries, operating the Service

Until the account is deleted (temporary retention is deleted after a short period). Backups until their retention period expires

Google LLC — Cloud Functions and Cloud Logging

Cloud Functions: United States. Cloud Logging: Google’s global infrastructure (including the United States)

When server functions are called / HTTPS

Reflection request content (not stored after processing), operation logs such as account identifier and crisis-signal screening information, request IP address, User-Agent, and time of request (default request logs)

Running server functions, diagnosing errors

Log retention period of 30 days (the retention period of the Cloud Logging default bucket)

Google LLC — Google Analytics

United States and other countries where Google operates data centers (Google’s global infrastructure)

As needed while you use the App / HTTPS

Usage events (classification values), app instance ID, Firebase Installation ID, iOS Identifier for Vendors (IDFV), device and app information, approximate location (the IP address is discarded after estimating location and before logging)

Usage statistics

Subject to the analytics data retention period set by the Company and Google’s service policies

PostHog Inc. — PostHog

United States

As needed while you use the App / HTTPS

Usage events, device and app information, identifier assigned by PostHog, approximate location estimated from the IP address

Usage statistics

Subject to the analytics data retention period set by the Company and PostHog’s service policies

Google LLC — Firebase App Check

United States and other countries where Google operates data centers (Google’s global infrastructure)

When requests are made to the server / HTTPS

App integrity check token

App integrity checks (not currently enforced by the server)

Tokens are valid for up to 7 days, and tokens used to prevent replay are retained for up to 30 days

Google LLC — Firebase Crashlytics

United States and other countries where Google operates data centers (Google’s global infrastructure)

When an app crash or error occurs / HTTPS

Error type, message, and stack trace, time of occurrence, device model, OS version, and app version, Crashlytics installation identifier (does not include original sentences)

Diagnosing errors, improving stability

90 days (Google policy)

Functional Software, Inc. — Sentry

United States

When an app crash or error occurs / HTTPS

Error type, message, and stack trace, time of occurrence, device model, OS version, and app version, build environment value, event identifier assigned by Sentry (does not include original sentences)

Diagnosing errors, improving stability

90 days (Sentry default retention policy)

Vercel Inc. — AI Gateway

United States and other regions operated by Vercel

At reflection requests and constellation completion / HTTPS

Today’s sentence, previous reflection conversation, (where recall is provided) the original text of sentences from entries you picked in weekly reviews, (once upon constellation completion) the sentences, dates, and tones of entries from that period and the constellation name, the generated reflection response (for response-rule checks)

Relaying AI requests

Not stored (zero data retention and training-prohibition settings applied). Operational metadata that does not contain sentence or conversation content (status, model, token usage, cost, duration, etc.) may be retained separately in accordance with Vercel’s operational policies

Claude model provider: Anthropic, PBC, or a Claude hosting provider to which the gateway routes the request

United States (gateway). The model hosting region depends on the provider’s settings (not fixed to the United States)

At reflection requests and constellation completion / HTTPS (via AI Gateway)

Today’s sentence, previous reflection conversation, (where recall is provided) the original text of sentences from entries you picked in weekly reviews, (once upon constellation completion) the sentences, dates, and tones of entries from that period and the constellation name

Reflection generation, crisis-signal classification, monthly comment generation

Not stored. Used only during processing under the zero data retention settings (training prohibition specified). Operational metadata that does not contain sentence or conversation content (status, model, token usage, cost, duration, etc.) may be retained separately in accordance with the provider’s operational policies

TypeSafe AI

United States (gateway). The processing region depends on the provider’s settings

Immediately after a reflection response is generated / HTTPS (via AI Gateway)

Text of the generated reflection response, today’s sentence, your messages in the most recent conversation (including quoted sentences where recall is provided), response language

Checking reflection responses for compliance with the conversation rules

Not stored. Used only during processing under the zero data retention settings (training prohibition specified)

Cloudflare, Inc. — Website provision, Web Analytics, launch notification storage

United States and other countries where Cloudflare operates its network. Launch notification sign-up information: Western United States (Cloudflare D1)

At Website visits and launch notification sign-ups / HTTPS

IP address, browser and device information, pages visited and referral source, launch notification sign-up information (email address, language, referral source, and the sign-up request’s IP address, User-Agent, and referring page)

Providing the Website, aggregating visit statistics, managing launch notification sign-ups

Visit information is subject to Cloudflare’s service policies. Launch notification sign-up information is subject to Section 9.1

The contact details of the recipients are as follows.

Recipient

Contact

Google LLC

googlekrsupport@google.com

PostHog Inc.

privacy@posthog.com

Functional Software, Inc.

privacy@sentry.io

Vercel Inc.

privacy@vercel.com

Anthropic, PBC

privacy@anthropic.com

Cloudflare, Inc.

privacyquestions@cloudflare.com

Transfers to Vercel, the Claude model provider, and TypeSafe AI take place only for users who have consented to AI processing, and account information (email, name, account identifier) is not included in the items transferred (Section 8.3). Because the Company does not fix the model provider or inference region in AI Gateway, the recipient and processing region in the Claude model provider row may vary from request to request, and zero data retention and training-prohibition settings apply whichever path is used (Section 8.2). Transfers to Google Analytics and PostHog take place only for users who have consented to usage statistics. Google Analytics, app integrity checks, crash reports, server operation logs (Cloud Logging), Vercel AI Gateway, the Claude model provider, and Cloudflare are services for which the provider does not limit the processing country to the United States, so their destination countries are stated as above.

13.3 Grounds for Cross-Border Transfers and How Consent Is Obtained

Category

Ground and method

Transfers essential to the Service (Google, Sentry, excluding usage statistics)

As outsourcing of processing and storage necessary to enter into and perform a contract, disclosed in this Policy (Article 28-8(1)(iii) of PIPA). Notice is given through the Terms of Service and this Policy before sign-in

Usage statistics transfers (Google Analytics, PostHog)

Processed on the basis of the data subject’s separate consent (Article 28-8(1)(i) of PIPA). Before obtaining consent, the Company informs you of the recipients, destination countries, items transferred, purposes of transfer, retention and use periods, how to refuse, and the effects of refusal

AI processing transfers (Vercel, Claude model provider, TypeSafe AI)

Processed on the basis of the data subject’s separate consent (Article 28-8(1)(i) of PIPA). Before obtaining consent, the Company informs you of the recipients, destination countries and regions, items transferred, purposes of transfer, retention and use criteria, how to refuse, and the effects of refusal

Website provision and launch notification storage (Cloudflare)

Disclosed in this Policy as outsourcing of processing and storage necessary to provide the Website and manage launch notification sign-ups

13.4 How to Refuse and the Effects of Refusal

Transfer

How to refuse

Effect of refusal

Google Firebase (authentication, database, server, app integrity checks, crash reports)

Do not use the Service, or delete your account

Because server storage is essential to providing the Service, you cannot use the Service

Sentry (crash reports)

Do not use the Service, or delete your account

Because crash reports are always active in release builds and there is no setting in the App to turn them off, you cannot use the Service

Google Analytics, PostHog

Do not consent on the usage statistics consent screen, or turn off “Send usage statistics” in the top bar under “Settings” > “Privacy”

No effect. You can use all features as they are

Vercel, Claude model provider, TypeSafe AI (AI reflections)

Choose “Not right now” on the AI processing consent screen, or turn off “Send my sentences to AI for reflections” in the top bar under “Settings” > “Privacy”

You cannot receive AI reflections, recall, crisis-signal screening and safety responses, or AI-generated monthly comments, and a fixed message is shown instead of a reflection. You can continue to use other features such as entries and star creation

13.5 Safeguards

  • Encryption in transit (HTTPS/TLS)
  • The items sent to AI processing providers are limited to sentences and conversation content, and account information (email, name, account identifier) is excluded
  • Training-prohibition and zero data retention (ZDR) settings are applied to all AI requests
  • AI requests are passed only through Vercel AI Gateway, and even when the gateway routes a request to another Claude hosting provider, the request is passed only to providers to which the zero data retention settings apply (Section 8.2)
  • Processors’ standard data processing terms apply

13.6 Rights Concerning Cross-Border Transfers

Users in the Republic of Korea may request access to information about cross-border transfers (Section 14) and may withdraw consent to AI processing transfers and usage statistics transfers at any time (Section 13.4). If you believe your rights or interests have been infringed in connection with a cross-border transfer, you may file a report with or seek advice from the organizations listed in Section 14.6.

14. Data Subject Rights and How to Exercise Them

14.1 Your Rights

Right

Description

Right of access

Access to the items of personal information processed, the purposes of processing, retention periods, and the status of outsourcing of processing and cross-border transfers

Right to request correction or deletion

Request to correct or delete inaccurate or incomplete personal information

Right to request suspension of processing

Request to suspend the processing of personal information (unless there are grounds for restriction under law)

Right to withdraw consent

Withdrawal of consent to processing based on consent, such as AI processing consent and usage statistics consent

Rights regarding automated decisions

Request an explanation of, and refuse, automated decisions (as stated in Section 8.8, the Company does not make automated decisions with significant effects)

Right to seek remedies for infringement

Report to or seek advice from the outside organizations listed in Section 14.6 in the event of an infringement of personal information

14.2 How to Exercise Your Rights

Method

Details

Account deletion (in the App)

Top bar “Settings” > “Account” > “Delete account”. After confirmation, your account data on the server and your account are deleted immediately (Section 9.2)

Withdrawal of AI processing consent (in the App)

Turn off “Send my sentences to AI for reflections” in the top bar under “Settings” > “Privacy”

Withdrawal of usage statistics consent (in the App)

Turn off “Send usage statistics” in the top bar under “Settings” > “Privacy”

Sign-out (in the App)

Top bar “Settings” > “Account” > “Sign out”. Server data (entries, reflection conversations, weekly reviews, etc.) is kept and only the account data on this device is deleted. Drafts in progress that existed only on this device and entries not yet uploaded to the server are lost and cannot be recovered

Email request

contact@passionfactory.ai (please include “StarWrite” in the subject line). Requests for access, correction or deletion of individual entries, suspension of processing, and account deletion when you cannot use the App are accepted

At present, the App does not offer a feature to edit or delete individual entries. If you would like an individual entry corrected or deleted, please send a request by email. At launch, a product feature for exporting entries to a file is not provided. However, this does not limit the rights recognized under applicable personal information protection laws, such as the rights of access, to receive a copy, to correction, and to deletion, and you may exercise those rights by email or other methods designated by the Company.

14.3 Identity Verification

When you request to exercise your rights by email, the Company may take necessary steps to verify that the requester is the account owner, such as replying to the email address of the account used to sign in. If you signed up using Apple’s “Hide My Email”, the Company’s reply sent to the private relay address may not reach you; in that case, the Company will verify your identity by other means. For requests made through an agent, the Company may request materials that confirm the authorization.

14.4 Processing Periods

Type of request

Processing period

Requests from users in the Republic of Korea for access, correction, deletion, or suspension of processing

Within 10 days from the date the request is received

Requests from users in Japan for disclosure, correction, suspension of use, etc.

Without delay

Requests from users in California, United States, to know, delete, or correct

Within the period prescribed by applicable law (in principle, within 45 days after receipt of the request)

14.5 Restrictions on Exercising Rights

If there are grounds prescribed by law, such as where retention is required by law or where there is a risk of unjustly infringing another person’s life, body, property, or other rights or interests, all or part of a request may be restricted; in that case, the Company will inform you of the reasons. In addition, the structure of the Service imposes the following restrictions.

Subject

Restriction

AI processing consent record

To prove consent, it is not deleted while the account exists but is kept in a withdrawn (“declined”) state, and it is deleted together with your account when you delete it

Reflection report documents

They cannot be viewed, corrected, or deleted in the App; please make a request by email

14.6 Remedies for Infringement of Rights (Users in the Republic of Korea)

Organization

Phone

Website

Personal Information Infringement Report Center (Korea Internet & Security Agency)

118 (no area code)

privacy.kisa.or.kr

Personal Information Dispute Mediation Committee

1833-6972

kopico.go.kr

Cyber Investigation Division, Supreme Prosecutors’ Office

1301 (no area code)

spo.go.kr

Cyber Investigation Bureau, Korean National Police Agency

182 (no area code)

ecrm.police.go.kr

15. Data Stored on Your Device and Automatic Collection Tools

15.1 Data Stored Locally on Your Device

So that StarWrite can be used offline, some data is kept in the App’s dedicated storage on your device.

Category

Content

Sent to server

Upon sign-out or account deletion

Entry cache

Device copy of entries stored on the server

The original is on the server

Deleted

Offline cache of server data

Device cache of Cloud Firestore (including reflection conversations, weekly reviews and segment reviews, constellations, age confirmation time, day boundary time, and AI processing consent record)

The original is on the server

Deleted

Drafts in progress

Sentences before submission, submission requests whose transmission was not completed

Drafts are not sent; submissions whose transmission was not completed are sent again at the next launch

Deleted

Usage statistics consent

Consent decision per account identifier

Not sent

Remains on the device (locked to the account identifier and not applied to other accounts)

Device settings

App language

The app language is sent with each reflection request (Section 8.3)

Remains on the device

When you delete the App, the data in the App’s dedicated storage is deleted with it. However, on iOS, sign-in information may remain in the device’s Keychain, so you may still be signed in after reinstalling. The Company cannot remotely delete data on your device.

15.2 Installation, Operation, and Refusal of Automatic Collection Tools

The mobile App automatically collects App usage information through the tools (SDKs) below. The Google Analytics SDK and the PostHog SDK operate only if you have consented to usage statistics, while Firebase App Check, Firebase Crashlytics, and Sentry operate regardless of consent. Analytics tools for the Website are described in Section 15.3.

Collection tool

Information collected

Purpose

Retention period

How to refuse

Google Analytics SDK

Usage events (classification values), screen views, app instance ID, Firebase Installation ID, iOS Identifier for Vendors (IDFV), device and app information, approximate location (Section 3.3)

Service improvement (usage statistics)

Subject to the analytics data retention period set by the Company and Google’s service policies

Turn off “Send usage statistics” in the top bar under “Settings” > “Privacy” (off by default; operates only if you consent)

PostHog SDK

Usage events, device and app information, identifier assigned by PostHog, approximate location (Section 3.3)

Service improvement (usage statistics)

Subject to the analytics data retention period set by the Company and PostHog’s service policies

Turn off “Send usage statistics” in the top bar under “Settings” > “Privacy” (off by default; operates only if you consent)

Firebase App Check

App integrity check token

Issued for security verification (not currently enforced by the server)

Tokens are valid for up to 7 days, and tokens used to prevent replay are retained for up to 30 days

There is no separate setting to refuse

Firebase Crashlytics SDK

Type, message, and stack trace of abnormal terminations and unhandled errors, time of occurrence, device model, OS and app version, Crashlytics installation identifier (Section 3.3)

Diagnosing errors, improving stability

90 days (Google policy)

Always active in release builds; there is no separate setting in the App to refuse

Sentry SDK

Type, message, and stack trace of abnormal terminations and unhandled errors, time of occurrence, device model, OS and app version, build environment value, event identifier assigned by Sentry (Section 3.3)

Diagnosing errors, improving stability

90 days (Sentry default retention policy)

Always active in release builds; there is no separate setting in the App to refuse

Each time the App launches, it downloads an app update information file from Firebase Hosting (Google LLC) to check whether a new version is available. This request does not contain account information or entries.

The Company does not use your App usage history for cross-context behavioral advertising, and does not collect the iOS Identifier for Advertisers (IDFA) or the Android Advertising ID (AAID). On iOS, the Identifier for Vendors (IDFV), which is not an advertising identifier, may be processed for usage statistics analysis.

15.3 Website Analytics and Cookies

The Website is provided through the services of Cloudflare, Inc., and visit statistics are aggregated only with Cloudflare Web Analytics. Cloudflare states that Web Analytics does not use cookies and does not identify individuals. The Website does not use other analytics tools such as Google Analytics or PostHog, and the Company does not send the original text of entries or AI conversations from the StarWrite App for web analytics.

The cookies used by the Website are as follows.

Cookie

Type

Purpose

Retention period

theme

Functional

Remembers the display theme (light or dark) chosen by the visitor

1 year

Visitors can refuse to store cookies or delete stored cookies in their browser settings, and can still use the Website even if they refuse cookies (in that case, the theme selection will not be remembered).

If the Company introduces new analytics or marketing tools on the Website that change how personal information is processed, the Company will update this Policy or a separate notice.

16. Other Grounds for Providing Personal Information

16.1 Provision Required by Law

The Company may provide personal information to the extent permitted by relevant laws in order to fulfill legal obligations or when investigative or supervisory authorities make a request through lawful procedures. In such cases, the Company will make efforts to inform you unless prohibited by law.

16.2 Business Transfers, Mergers, and Similar Transactions

If the Company transfers the Service through a merger, business transfer, or the like, your personal information may be transferred along with it. In that case, the Company will, in accordance with relevant laws, notify you in advance of the transfer, the recipient, and what you can do if you do not want your information transferred, and will ensure that the same level of protection continues to apply after the transfer.

17. Additional Regional Information

17.1 Users in the Republic of Korea

This Policy has been prepared on the basis of PIPA and related laws, and all of its provisions apply to users in the Republic of Korea. The personal information of users in the Republic of Korea is stored and processed mainly in the United States in accordance with Section 13, and usage statistics in Google Analytics, app integrity checks, crash reports, server operation logs, and AI processing may also be processed in regions outside the United States.

17.2 Users in Japan (Additional Information Under APPI)

Item

Details

Purposes of use

Personal information is used only for the purposes published in Section 5

Provision to a third party in a foreign country — Google LLC

Google LLC, located in the United States, processes personal information for sign-in, entry storage, server functions, app integrity checks, crash reports, and (if you consent) usage statistics (Section 13, Appendix A.1). Sign-in, entry storage, and server functions are processed in the United States, and usage statistics, app integrity checks, crash reports, and server operation logs are processed on Google’s global infrastructure (including the United States). This provision is made on the premise that the recipient has established a system for continuously taking measures equivalent to those required under APPI (Article 28(1)), and upon your request, the Company will provide information on those measures (Article 28(3))

Provision to a third party in a foreign country — PostHog Inc.

If you consent to usage statistics, PostHog Inc. (United States) processes personal information in PostHog Cloud (United States) (Section 13, Appendix A.6). This provision is made on the premise that the recipient has established a system for continuously taking measures equivalent to those required under APPI (Article 28(1)), and upon your request, the Company will provide information on those measures (Article 28(3))

Provision to a third party in a foreign country — Functional Software, Inc.

For crash reports, Functional Software, Inc. (United States) processes personal information in Sentry (United States) (Section 13, Appendix A.7). Crash reports are always sent in release builds regardless of consent to usage statistics. This provision is made on the premise that the recipient has established a system for continuously taking measures equivalent to those required under APPI (Article 28(1)), and upon your request, the Company will provide information on those measures (Article 28(3))

Provision to a third party in a foreign country — Vercel Inc. (United States), the Claude model provider (Anthropic, PBC (United States), or a Claude hosting provider to which the gateway routes the request), TypeSafe AI

AI processing takes place only if you give consent in the App (consent under Article 28(1)). Vercel AI Gateway may relay requests in multiple Vercel regions, including the United States, and the processing region of the Claude model is not fixed to the United States and depends on the provider’s settings (Sections 8.2 and 13.2). Before obtaining consent, the Company provides information on the names of the destination countries, the personal information protection systems of those countries, and the measures taken by the recipients (Article 28(2))

Measures by processors

The Company oversees its processors so that they continuously take measures equivalent to those required under APPI, and contractual measures are governed by the data processing terms that apply between the Company and each processor

Requests concerning retained personal data

You may request notification of the purposes of use, disclosure, correction, addition, or deletion, suspension of use or erasure, and suspension of provision to third parties, in the manner described in Section 14. Disclosure will be made by the method you designate, including electronic means

Security control measures

The measures in Section 11 apply

Supervisory authority

Personal Information Protection Commission of Japan (個人情報保護委員会)

17.3 Users in California, United States (Additional Information Under the CCPA/CPRA)

This Section applies to users residing in California to the extent the CCPA applies. The Company discloses the categories of personal information below only to the service providers stated in the table and only for business purposes, and the same was true for the preceding 12 months.

Category of personal information collected (statutory category)

Applicable items

Source

Purpose

Disclosure for a business purpose and categories of recipients

Identifiers

Email address, display name (Google sign-in), account identifier, app instance ID, Firebase Installation ID, Crashlytics installation identifier, event identifier assigned by Sentry, identifier assigned by PostHog, iOS Identifier for Vendors (IDFV), IP address of sign-in and server requests

You, sign-in providers, device

Account management, providing the Service, usage statistics (with consent), security and error diagnosis

Disclosed — service providers (cloud hosting, analytics (with consent))

Internet or other electronic network activity information

App usage events, screen views, device model, OS and app version, crash reports, server operation logs, request information (User-Agent)

Device, server

Service improvement (usage statistics with consent), security and error diagnosis

Disclosed — service providers (cloud hosting, analytics (with consent))

Geolocation data (not precise location)

City-level location estimated from the IP address

Device

Usage statistics (with consent)

Disclosed — service providers (analytics (with consent))

Audio, electronic, visual, or similar information

Entry sentences, reflection conversations, reflection reports (electronic text)

You

Providing the Service, AI reflections (with consent)

Disclosed — service providers (cloud hosting, AI processing (with consent))

Sensitive personal information

Health- or mental-health-related content that you voluntarily leave in sentences

You

Providing the Service; (if you consent to AI reflections) providing safety responses based on crisis-signal screening (Section 7.3). Not used to infer your characteristics

Disclosed — service providers (cloud hosting, AI processing (with consent))

Item

Details

Sale and sharing

The Company does not sell personal information or share it for cross-context behavioral advertising, and has not done so in the preceding 12 months

Under 16

The Company does not sell or share the personal information of consumers under 16 years of age

Limiting the use of sensitive personal information

Because the Company does not use sensitive personal information for purposes other than those permitted by the CCPA (such as providing the Service), it does not have a separate procedure for requests to limit its use

Your rights

Right to know (the categories of personal information collected, the categories of sources, the purposes of collection, the categories of third parties to whom it was disclosed, and a copy of the specific pieces of personal information collected), right to delete, right to correct, and the right not to be discriminated against for exercising your rights

Right to opt out of sale and sharing

Because the Company does not sell or share personal information, it does not have a separate opt-out procedure. If the Company begins to sell or share personal information, it will first update this Policy and explain how to opt out

How to make a request

By email (contact@passionfactory.ai) or through account deletion in the App (Section 14.2). Identity verification is governed by Section 14.3, and an authorized agent may also make a request. Processing periods are governed by Section 14.4

Retention periods

Governed by Section 9

Updates to this section

This Section is reviewed and updated at least every 12 months

The principles of this Policy (no sale of personal information, acceptance of deletion and correction requests) also apply to users residing in other U.S. states, and if the laws of the relevant state guarantee more favorable rights, those laws will be followed.

17.4 Other Regions

StarWrite is a global service, and this Policy also applies when you use the Service in regions not specifically addressed above. If the mandatory laws of the relevant region guarantee you more favorable rights, those laws prevail.

17.5 EU/EEA and the United Kingdom

EU/EEA member states and the United Kingdom are not currently official target regions of the Service, and the Company has not designated a representative under Article 27 of the GDPR or Article 27 of the UK GDPR. This Policy does not represent that the Company has satisfied all requirements of the laws of those regions.

18. Minimum Age and Protection of Children’s Personal Information

18.1 Minimum Age

The minimum age for using the Service is 16. Right after Google or Apple sign-in, the App asks you to confirm that you are 16 or older; if you answer that you are under 16, you cannot use the Service and are signed out. Accordingly, use by children under 14 protected by Article 22-2 of PIPA and by children under 13 protected by the U.S. Children’s Online Privacy Protection Act (COPPA) is also not permitted.

18.2 Guardian Consent

Persons under 16 cannot sign up.

18.3 Guardian Access to Entries

The Company does not provide guardians with any feature or permission to automatically view your entries. If a legal representative exercises rights under relevant laws, the Company will handle the matter individually in accordance with the requirements and procedures prescribed by those laws.

18.4 Information of Persons Under the Minimum Age

Because age confirmation relies on your answer, the possibility that information of persons under 16 may be collected cannot be completely ruled out. In particular, even if you answer that you are under 16, the Firebase Authentication account information created during sign-in (account identifier, email address, display name for Google sign-in, etc.) is not deleted merely by signing out. Because usage statistics are collected only if you consent after age confirmation, usage statistics are not collected from persons who answer that they are under 16. If the Company becomes aware that personal information of a person under 16 has been collected, it will delete that information without delay. A legal representative who believes that their child’s personal information has been collected should contact contact@passionfactory.ai.

19. Changes to This Policy

19.1 Notice of Changes

The Company may change this Policy in response to changes in laws, processors, or Service features. When it does so, the Company will give notice before the changes take effect in accordance with relevant laws, and will obtain consent for any change that requires consent under law (e.g., a change in AI processing providers or in the items sent).

19.2 Change History

Version

Date

Details

v1.5

Official service launch date

First effective version

Previous versions of this Policy are available upon request to contact@passionfactory.ai.

20. Contact Information

Item

Details

Chief Privacy Officer

Jeonghoon Lee (CEO)

Email

contact@passionfactory.ai

Email subject

Please include “StarWrite” in the subject line (recommended). Emails with a different subject line will also be received

Phone

+82-70-7756-1070

Address

161 Hakdong-ro, Gangnam-gu, Seoul, Republic of Korea

For organizations where users in the Republic of Korea can report infringements of personal information, please see Section 14.6.

Appendix A: Processors and Cross-Border Transfers in Detail

Reference Date for technical information: the effective date of this Policy

This Appendix sets out the processors, processing locations, and items processed as of the Reference Date. Information that is not appropriate to disclose for security reasons or that changes frequently, such as internal operational identifiers and configuration values, is not included. The content below reflects the configuration as of the Reference Date and may change depending on future infrastructure configuration. If it changes, the Company will update this Policy and notify you, and will follow the applicable procedures for any change that requires consent or notice under law.

A.1 Google LLC — Firebase / Google Cloud

Service

Use

Processing location

Items processed

Retention

Cross-border transfer

Firebase Authentication

Sign-in authentication, account management

United States

Account identifier, email address, display name and profile photo URL (Google sign-in only), IP address of sign-in requests

Until the account is deleted (after deletion, removed from operational and backup systems within 180 days). IP addresses are retained for several weeks (as published by Firebase)

Yes

Cloud Firestore

Storing entries, reflection conversations, consent records, weekly reviews, constellations, account settings, and reports; temporary retention of reflection responses; call count records; account deletion markers

United States

Items stored on the server under Sections 3.1 to 3.3, 8.6, and 8.10

Deleted immediately upon account deletion. Temporary retention is deleted after a short period. Account deletion markers for up to 1 hour. Backups are destroyed after their retention period expires

Yes

Cloud Functions

Relaying reflections, crisis-signal screening, monthly comment generation, receiving reports, account deletion

United States

Reflection request content (not stored after processing), account identifier, device country code and app language

Only while the request is being processed

Yes

Cloud Logging

Server operation logs, default request logs

Google’s global infrastructure (including the United States)

Account identifier, function name, reason and response type of received reflection reports, crisis-signal screening information (keyword rule level, AI classification result, classification method, region), request IP address, User-Agent, and time of request. Logs written by the Company do not record original sentences or response text, but in exceptional errors, such as malformed requests, the request content may remain in the server framework’s error logs

30 days (the retention period of the Cloud Logging default bucket)

Yes

Google Analytics

Usage statistics (if you consent)

Google’s global infrastructure (including the United States)

Usage events (classification values), screen views, app instance ID, Firebase Installation ID, iOS Identifier for Vendors (IDFV), device and app information, approximate location (the IP address is discarded after estimating location and before logging) (no account identifier set)

Subject to the analytics data retention period set by the Company and Google’s service policies

Yes

Firebase App Check

Issuing app integrity check tokens (not currently enforced by the server)

Google’s global infrastructure (including the United States)

App integrity check token (device attestation data is passed to Apple App Attest and DeviceCheck or Google Play Integrity, and App Check does not retain it)

Tokens are valid for up to 7 days, and tokens used to prevent replay are retained for up to 30 days

Yes

Firebase Crashlytics

Crash reports (always active in release builds)

Google’s global infrastructure (including the United States)

Error type, message, and stack trace, time of occurrence, device model, OS version, and app version, Crashlytics installation identifier (no account identifier set; does not include original sentences)

90 days (Google policy)

Yes

Contractual and processing terms are governed by the data processing terms that apply between the Company and each processor.

A.2 Vercel Inc. — AI Gateway

Item

Details

Provider

Vercel Inc. (United States)

Role

Relaying AI requests between the StarWrite server and AI model providers

Scope

Reflection requests, crisis-signal classification requests, monthly comment generation requests, and response-rule check requests of users who have consented to AI processing

Items processed

Today’s sentence, previous reflection conversation, (where recall is provided) the original text of sentences from entries you picked in weekly reviews, (once when a constellation is completed) the sentences, dates, and tones of entries from that period and the constellation name (Section 8.10), and, for response-rule checks, the generated reflection response and the response language

Items not processed

Account information such as email, name, account identifier, and IP address. However, names and other details that you yourself write in a sentence are passed on as they are

Processing location

United States and other regions operated by Vercel

Training use

An option prohibiting training on prompts is specified for each request. Vercel does not use prompts for training

Provider selection

The Company does not fix the model provider or inference region. The gateway may route requests to another hosting provider that offers the same Claude model (e.g., Claude hosted by a cloud provider), and, in accordance with the zero data retention settings specified for each request, passes requests only to providers to which zero data retention applies

Retention

Prompts and responses are not stored (zero data retention settings applied). Operational metadata that does not contain sentence or conversation content (status, model, token usage, cost, duration, etc.) may be retained separately in accordance with Vercel’s operational policies

Cross-border transfer

Yes (based on in-app AI processing consent)

A.3 Claude Model Provider — AI Models

Item

Details

Provider

Anthropic, PBC (United States), or a Claude hosting provider to which Vercel AI Gateway routes the request (such as Claude hosted by a cloud provider)

Models

Claude family (a model for reflection generation and monthly comment generation, and a model for crisis-signal classification). If the model or provider configuration changes, this Appendix will be updated

Role

Generating reflection responses, classifying crisis signals, generating monthly comments

Items processed

The items processed under A.2 that are passed through Vercel AI Gateway (excluding response-rule check requests)

Processing location

Regions operated by the provider, including the United States. The processing region is not fixed to the United States and depends on the provider’s settings

Training use

Training prohibition is specified for each request through AI Gateway. Anthropic’s published default policy is not to use inputs and outputs of its commercial API for model training

Retention

Zero data retention (ZDR) settings apply whichever provider processes the request. The sentences and conversations sent are used only during processing and are not stored by the provider. Operational metadata that does not contain sentence or conversation content (status, model, token usage, cost, duration, etc.) may be retained separately in accordance with the provider’s operational policies

Cross-border transfer

Yes (based on in-app AI processing consent)

A.4 TypeSafe AI — Response-Rule Check Model

Item

Details

Provider

TypeSafe AI (a zero data retention provider registered with Vercel AI Gateway)

Model

Jev (a judgment model). If the model or provider changes, this Appendix will be updated

Role

Checking, rule by rule, whether a generated reflection response followed the Company’s conversation rules (no questions, no advice, etc.)

Scope

Reflection responses for users who have consented to AI processing

Items processed

Text of the generated reflection response, today’s sentence, your messages in the most recent conversation (including quoted sentences where recall is provided), response language. Account information is not sent

Processing location

Via Vercel AI Gateway. The processing region depends on the provider’s settings

Training use

Training prohibition is specified for each request through AI Gateway

Retention

Zero data retention (ZDR) settings applied. The sentences and responses sent are used only during processing and are not stored by the provider

Use of check results

Used only for the server to decide whether to send the response as is. Server operation logs record only the names of the rules violated and the check method, not the response text or sentences

Development environment

In the Company’s development environment, the same provider is also used for crisis-signal classification (Section 7.3). In the production environment, crisis-signal classification is handled by the Claude model provider in A.3

Cross-border transfer

Yes (based on in-app AI processing consent)

A.5 Sign-In Providers

Provider

Role

Processing location

Items the Company receives

Notes

Apple Inc. (Sign in with Apple)

Performs authentication only

Subject to the provider’s policies (provider located in the United States)

Apple user ID, email address (may be a private relay address). The App may receive your name, but the Company does not store it on its servers

Upon account deletion, the Company revokes the Apple token. Apple’s privacy policy also applies

Google LLC (Google Sign-In)

Performs authentication only

Subject to the provider’s policies (provider located in the United States)

Google user ID, email address, display name and profile photo URL

Upon account deletion, the Company requests that the App’s connection to your Google account be removed. Google’s privacy policy also applies

A.6 PostHog Inc. — Usage Statistics

Item

Details

Provider

PostHog Inc. (United States)

Role

Aggregating and analyzing usage statistics for the App

Items processed

Usage events, device and app information, identifier assigned by PostHog, approximate location estimated from the IP address. Original sentences and reflection conversations are not sent

Scope

App users who have consented to usage statistics

Processing location

PostHog Cloud (United States)

Retention

Subject to the analytics data retention period set by the Company and PostHog’s service policies

How to refuse

Do not consent, or turn off “Send usage statistics” in the top bar under “Settings” > “Privacy”

Cross-border transfer

Yes

A.7 Functional Software, Inc. — Sentry (Crash Reports)

Item

Details

Provider

Functional Software, Inc. (operator of Sentry, United States)

Role

Collecting and analyzing app crash reports

Items processed

Type, message, and stack trace of abnormal terminations and unhandled errors, time of occurrence, device model, OS version, and app version, build environment value (dev or prod), event identifier assigned by Sentry. No account identifier is set, the App is configured not to store the IP address as a user attribute, and screenshots and the App’s log output are not sent. Original sentences and reflection conversations are not sent. However, if an error message includes a server document path, it may contain an account identifier

Scope

Users of release builds of the App (regardless of consent to usage statistics)

Processing location

Sentry US region (United States)

Retention

90 days (Sentry default retention policy). Because no account identifier is set, reports are not deleted together with the account upon account deletion

How to refuse

None (always active in release builds; no setting in the App to turn it off)

Cross-border transfer

Yes

A.8 Cloudflare, Inc. — Website

Item

Details

Provider

Cloudflare, Inc. (United States)

Role

Providing the Website (starwrite.me) (hosting and delivery), aggregating visit statistics (Cloudflare Web Analytics, no cookies), storing launch notification sign-up information (Cloudflare D1)

Scope

Website visitors and people who sign up for launch notifications. Not involved in the mobile App’s entries or AI processing

Items processed

IP address, browser and device information, pages visited and referral source. For launch notification sign-ups, the email address, language, and referral source, and the sign-up request’s IP address, User-Agent, and referring page

Processing location

Cloudflare’s global network (including the United States). Launch notification sign-up information: Western United States

Retention

Visit information is subject to Cloudflare’s service policies. Launch notification sign-up information is deleted once the purpose of the launch announcement has been achieved or upon a deletion request (Section 9.1)

Cross-border transfer

Yes

A.9 Tools Not Used

As of the Reference Date, the Company does not use ad-serving SDKs, push notifications (FCM), or payment processing tools in the mobile App. Even though Google Analytics is used, it is configured not to collect the Android Advertising ID (AAID) or the iOS Identifier for Advertisers (IDFA). Non-advertising analytics identifiers, such as the iOS Identifier for Vendors (IDFV), may be processed within the scope of Sections 3.3 and 15.2. Launch notifications and analytics tools for the Website are described in Sections 3.7 and 15.3. If the Company introduces new tools that change how personal information is processed, it will update this Policy first.

Appendix B: Purposes for Which the Company Does Not Use Entry Content

Purpose not pursued

Details

Advertising

The Company currently does not display advertising. If the Company introduces advertising in the future, it will update this Policy and obtain any consent required by law before doing so

AI training

What you enter is not used for training, retraining, or additional training (fine-tuning) of AI models or for building evaluation datasets, and training-prohibition and zero data retention settings are applied to AI processing providers

Sale

Entries and personal information are not sold to third parties

Profiling

Entries are not used to create profiles that infer your tendencies, emotional state, or health condition. Crisis-signal screening (Section 7.3) is a safety measure concerning a single sentence, and the monthly comment (Section 8.10) puts into words the tone statistics of that month’s entries; neither is a profile that infers your characteristics

Analytics events

Original sentences, drafts in progress, and reflection conversations are not included in usage statistics

Disclosure

Except as required by law (Section 16.1), entries are not shown or disclosed to other people without your consent. A constellation share card is an image that you yourself create and export, and it does not contain original sentences (Section 8.10)

Operator viewing

Officers and employees are prohibited from viewing entries in the ordinary course. Authorized personnel access entries only to the minimum extent necessary and only in very limited cases, such as at your request (including reviewing reports and handling requests to exercise your rights), responding to outages or security incidents, or fulfilling legal obligations. This is a control based on operational policy, and entries are not end-to-end encrypted (Section 11.1)

End of This Policy

This Policy is provided in Korean and English. For users in the Republic of Korea, the Korean version prevails, and for users in other regions where the Service is available, the English version prevails. However, if the mandatory laws of the country or region where you reside provide otherwise, those laws prevail.