원문 언어로 표시됩니다.
Myakuren Privacy Policy
Last updated: 2026-09-26
1. Introduction
NITI Technology Inc. (“we”, “us”) operates “Myakuren” (the “Service”), an AI-based communication practice tool for dating and relationship skills. This policy explains how we handle information about users and about third parties (practice partners) that users may input.
2. Information We Collect
| Category | Content | Retention |
|---|---|---|
| Account information | Email address, optional Google sign-in identifier, date of birth | Stored (encrypted) |
| Payment information | Stripe-managed reference ID (we do not hold card numbers) | Reference ID only |
| Practice records | Session logs, scoring results, “interest level” diagnosis results, affinity state with characters | Stored (encrypted) |
| Uploaded images | Screenshots attached for diagnosis or partner-profile features | Never stored |
| Raw text of imported chat history | Pasted text | Discarded immediately after feature extraction |
| Partner profile | Extracted traits only (age range, interests, reply-length tendencies, etc.) | Stored (encrypted). Names/contacts excluded |
| Usage data | Access logs, error logs, aggregate cost-monitoring data | Stored; only identifiers, not values |
3. We Do Not Store Uploaded Images
We do not store uploaded images, under any plan, any feature, or regardless of user consent. Images are processed in memory only and are never written to storage, logs, or backups.
Any call to an external AI model uses a logging-disabled routing path (AI Gateway’s no-log route) and only providers that contractually exclude our input from their model training.
Before upload, your device blurs the likely name/avatar area of the screenshot locally. If you prefer not to send an image at all, a text-paste path is always available.
4. Raw Text of Imported Chat History
Once speaker attribution and feature extraction are complete, we discard the raw text immediately. Our database has no column for storing raw imported text.
Flow:
- Paste chat history
- The app automatically attributes lines to “you” / “them” and displays the result
- You review and correct the attribution (this step cannot be skipped)
- Once confirmed, features are extracted
- The raw text is discarded
5. Automatic Masking of Third-Party (Practice Partner) Personal Information
The practice partner’s name, photo, contact details, employer, and social-media handles are automatically masked through three layers:
- Layer 1 (deterministic rules): phone numbers, emails, URLs, social handles, address patterns, company/school names, dates of birth, etc.
- Layer 2 (model-based extraction): context-dependent names, places, and nicknames, via a model called through the no-log routing path.
- Layer 3 (user review): you review the masked result and can add masks the system missed.
Immediately before storage, Layer 1’s deterministic rules are re-applied as a final check; if anything is still detected, the record is not saved.
6. What We Store vs. What We Never Store
Only extracted traits are stored in a partner profile.
- Stored: age range, interests, reply-length tendencies, tone patterns, a summary of stated interests
- Never stored: real name, social handles, employer, city of residence, photos, or quoted raw text
7. Use for Model Training
- Your practice partner’s imported chat history or images are never used to train our models, regardless of consent.
- Your own conversation logs are used for training only with your explicit consent. Declining this consent does not restrict access to any feature.
8. Privacy Modes
| Mode | Processing | Available in | What we can promise |
|---|---|---|---|
| Anonymized / no-log | Proper nouns are replaced before processing; no conversation log is stored | All scenes | No conversation record remains; no personally identifying data is stored |
| Standard | Conversation log stored | All scenes | Enables growth reports and continuous scenarios |
9. Viewing, Editing, Deleting, and Exporting Your Data
You can view, edit, delete, and export your practice history, character relationships, and partner profiles at any time. These functions are never limited to paid plans.
10. Deletion on Account Closure
When you close your account, we delete all associated data except where retention is legally required. Deletion covers D1 (database), R2 (voice cache, diagnosis cards), KV (usage counters), Vectorize (conversation embeddings), and Durable Objects (in-progress sessions). We also delete or anonymize the corresponding Stripe customer record.
11. Crisis-Detection Records
Records created when suicidal ideation or severe self-deprecation is detected are treated as our most sensitive data category, stored encrypted, and used only in aggregate — never reviewed individually except when responding to the incident itself. Crisis-support resources shown are specific to your region.
12. Disclosure to Third Parties
We do not share your personal information with third parties except:
- With your consent
- As required by law
- To service providers necessary to operate the Service (payment processing via Stripe; AI model calls via AI Gateway providers), under appropriate data-processing agreements
The lawfulness of handling third-party (practice partner) information, and consistency with major dating-app terms of service, has not yet been confirmed by outside counsel.
13. Cross-Border Data Transfers
For EU users, we are technically verifying whether data can be kept within the EU. We will not make the Service generally available to EU users until this verification is complete. Whether AI-model calls constitute a cross-border transfer will be documented once that verification is complete.
14. Disclosure That You Are Talking to an AI
We disclose, on-screen and in every region, that all character interactions within the Service are AI-generated and not conversations with a real person (in support of EU AI Act Article 50 transparency obligations and relevant US state laws).
15. Contact
- Company: NITI Technology Inc.
- Address: 3F Sobunkan Bldg., 6-25-14 Hongo, Bunkyo-ku, Tokyo, Japan
- Data protection contact: 金子友哉 (Yuya Kaneko)
- Contact channel: support@myakuren.com
16. Region-Specific Addenda
Addenda for the EU, UK, US, South Korea, and Taiwan (GDPR legal basis and erasure-right coverage, UK GDPR, US state laws / COPPA, Korean PIPA, Taiwan PDPA) will be added once findings are confirmed by outside counsel. We will not publish the Service in a language/region until that region’s legal review is complete.
17. Changes to This Policy
We may revise this policy to reflect legal or product changes. Material changes will be notified to users in advance.