Privacy Policy

• Google sign-in profile data such as email address, display name, and profile image.
• Guest or account-linked game progress, replay records, settings, and cloud-save data.
• Dialogue replies you type into training scenes and AI-generated evaluation results.
• Birth date, display name, exploration focus, Kin notes, AI mirror conversations, and mirror-written note drafts entered in Fathom.
• Feedback or problem reports you submit inside the app.
• Basic technical information needed to run the service, such as browser requests, timestamps, and error logs.

• To let you sign in, continue a journey, and sync practice progress when cloud save is enabled.
• To evaluate training replies, generate NPC reactions, provide coaching feedback, or generate AI mirror replies and note drafts in Fathom.
• To diagnose bugs, broken AI generations, login problems, or save issues.
• To improve the app's training design and scenario quality.

• It processes: public account identifiers (Page name, handle, numeric ID), the text and images we prepare for publication, and the identifiers and links of published results.
• It reads aggregate performance figures, such as view counts, reach, and the number of likes and comments.
• When these accounts receive comments or direct messages, it may read the content of those comments and messages, along with the sender's display name and account identifier, in order to reply, provide support, and manage the community. Where AI helps draft or send a reply, that text may be sent to AI service providers for processing (see AI Processing above).
• That content is used only to operate and respond on that account. We do not use it to build audience lists, we do not perform cross-account individual profiling, we do not sell it, and we do not share it with third parties beyond advertising we run for the account itself.
• We never write comment content or sender identities into any public record, including public code repositories and external documents.
• It does not proactively collect personal data about ordinary Facebook or Instagram users who have not interacted with us.

The credentials needed to access these accounts are stored encrypted inside the company, used only at the moment of publishing or reading performance figures, and are never written into documents, conversation logs, or shared with any third party. We do not sell or trade this data, and we do not use it for third-party purposes beyond advertising we run for the account itself. For client accounts we operate, data and access are returned or revoked when the engagement ends.

After explicit authorization, the OAuth connection page processes the administrator's display name, manageable Pages, recent post metadata, and public comment content, timestamps, comment IDs, sender names, and sender IDs inside the current browser tab. OAuth tokens and API responses are not sent to or retained by this website's server. The token exists only in page memory and disappears when the page is refreshed, closed, or cleared. The page does not publish, reply, edit, delete, moderate, or create advertising audiences.

• When you add one of these accounts as a friend, LINE gives us a user identifier that represents you. LINE issues this identifier per provider, and the accounts we operate are all registered under the same provider: if the same person adds different accounts among them, the identifier is the same, so we may be able to recognize that it is the same friend. We never guess who you are from your name, profile picture, or behavior.
• We read aggregate account statistics, such as friend counts and message delivery and open counts.
• We may send messages to friends through the API, such as broadcast messages or rich menus.
• One-to-one chats with 宇宙媽媽賞 are still handled by the account operator in LINE's own account management tool. Our systems do not receive the content of its one-to-one messages. If that changes, we will update this policy first.
• We currently do not combine records of the same friend across different accounts. If we decide to, we will update this policy first.
• Friend data is used only to operate and respond on that account. We do not sell or trade it, and we never write it into any public record.

One-to-one chats with 四分之三 Three-Quarters: answered by our AI digital human FoZone (佛佐), and kept

• Who collects: Three-Quarters International Ltd.
• Why: to reply to your messages, to continue the conversation between you and FoZone (she remembers your name and what the two of you talked about), to maintain our contact with you, and to let FoZone learn from conversations.
• What: your LINE user identifier, your display name at the time (not your profile picture), the messages, images, videos, audio and files you send, and the replies our system sends you. Darren, who is responsible for the account, may also reply to you himself. Messages he sends through our system are labeled “Darren” above the message and recorded apart from FoZone's replies. Replies he types inside LINE's account management tool do not reach the system; he adds that part of the conversation afterwards, also marked as his own words.
• How: FoZone is an AI, played by an AI model; each reply is labeled with the model in use (currently Gemini on Google Cloud Vertex AI). To generate a reply, the system gives the model only the exchanges between you, FoZone and Darren, never anyone else's conversations, plus FoZone's own weekly learning notes, which, apart from Darren himself, contain no one's name and quote no one's words. In your first conversation, FoZone tells you she is an AI and that the conversation is kept. For quotes, contracts, payments, scheduling, requests to see or delete your data, when a human is needed, or if you mention danger, she hands the conversation to Darren, who is responsible for it. The system notifies Darren through Telegram; the notification contains only your display name, an internal code and the reason, not your messages. When you ask about the Three Realms Protocol, a term in it, or a document code (such as LEX·007), or follow up on protocol material you were just discussing, the system may first look up relevant passages in the Three Realms Protocol public library, a lookup service we operate, and then give those limited excerpts, together with the conversation, to the model to compose the reply. The lookup receives only one search phrase of at most 240 characters: for an explicit question, it is derived from your latest text message; for a follow-up, FoZone writes the topic to look up from the conversation, such as a document code or a term, so it may draw on your earlier messages. The system removes your LINE display name; the search phrase never includes your LINE user identifier, and the full conversation is never sent. The library is “public” in the sense that anyone can read its contents; the query you send is not made public.
• Who else, and where: inside the company, only Darren has access. Service providers processing it on our behalf are Google (Vertex AI to generate replies and, once a week, to turn conversations into learning notes without friends' names; the company's Google Drive for backups, accessible only inside the company), Cloudflare (an encrypted tunnel that carries LINE's messages to our machine), Telegram (notifications to Darren), and Anthropic and OpenAI (if we switch to their models to generate replies; we currently use Google); LINE itself also keeps your chat history. The Three Realms Protocol library lookup service is operated by us on Google Cloud, currently in Google's Belgium region (europe-west1); the service and its entry point do not log the content of queries, only technical records such as request time, path and outcome. These services may process data outside Taiwan. FoZone learns from conversations: we may turn them into learning notes that do not contain your name, and use those notes to adjust FoZone. Under their commercial terms, the model providers do not use these conversations to train their own models. Apart from the service providers listed above, we do not give conversations to any other third party.
• How long: until you ask us to delete them or the account stops operating.
• Your rights: under Article 3 of Taiwan's Personal Data Protection Act you may ask to see or review your data, get a copy, supplement or correct it, have us stop collecting, processing or using it, and have it deleted. Email [email protected], or just tell FoZone and she will pass it to Darren. Once you ask us to stop collecting, the system no longer records or replies to messages you send, until you ask us to resume; you can also ask us to delete what is already stored. When we delete, we remove your records from the system database, saved media files and cloud backups, and retract notifications to Darren sent within the last 48 hours. Content already turned into FoZone's learning notes contains no name and cannot be traced back to you, so it falls outside deletion. The chat history on LINE's side is kept by LINE, and you can delete the chat room in LINE.
• If you don't provide it: whether to add the account or send messages is up to you. Without messages we cannot reply to you on LINE; you can email us instead.

As with the tool above, the credentials needed to access these accounts are stored encrypted inside the company and unlocked only at the moment of use. For client accounts we operate, we stop sending messages or reading data through the account when the engagement ends.