This service is an experimental prototype designed for self-care reflection and thought-sorting. It is not medical software, therapy, clinical diagnosis, or emergency support. If you are experiencing acute distress, self-harm thoughts, or an emergency, stop using the service immediately and contact your local healthcare provider or emergency services.
Reflections on saved conversations
Only when you press Generate review on My Page, Gemini processes saved, completed conversations in the same language from the selected last 1–30 days. It produces a self-care review and working hypothesis. The hypothesis is not a diagnosis and is not automatically used in regular chats. Unsaved turns and text truncated when saved are excluded.
We mask detectable personal identifiers again in inputs and outputs, but cannot guarantee removal of names, employers, or full addresses. Provider retention depends on the model backend and its terms. ArigatoAI does not include review text in operational logs or traces.
Reviews, aggregate metadata, sources, and the current working hypothesis are stored under your UID in Cloud Firestore, without an automatic deletion deadline. Only you can view or delete them. Review deletion, deletion of a source conversation, account deletion, and identity-verified erasure delete the relevant data. Deleting the latest review also deletes the current working hypothesis, which is rebuilt next time.
After account deletion or erasure, we retain only a UID-linked boolean erasure marker without an automatic deletion deadline to prevent already authenticated, delayed requests from recreating data. It contains no conversation text, name, or email.
1. Core Principles
The Service provides AI-assisted reflective dialogue informed by cognitive behavioral therapy (CBT) and Adlerian psychology principles. Guided by Privacy-by-Design, we avoid unnecessary data collection and process only what is needed to deliver a helpful, secure coaching experience.
2. Information We Collect
We collect and process the following categories of information:
- Dialogue Inputs: Text messages, emotional check-in scores, and reflection responses you submit during coaching sessions.
- Audio Input Data (Optional): Voice recordings submitted when using the microphone transcription feature (only after your explicit prior consent).
- Account Information (Upon Sign-in): User ID (UID), email address, and display name provided through Google Sign-In.
- Technical Information: Session cookies, IP address (temporarily referenced for rate limiting and abuse prevention), and browser user-agent.
3. How We Use Information
We use collected information solely for the following purposes:
- Delivering interactive AI coaching, cognitive defusion, and Socratic reflection.
- Transcribing voice inputs into editable text in real time.
- Enabling signed-in users to review past consultations and resume sessions via My Page.
- Protecting the service from automated abuse, DDoS attacks, and prompt injection attacks.
- Ensuring technical stability and diagnosing service errors (using sanitized metadata logs only).
4. Audio Input and Gemini API Processing
When you use the optional voice transcription feature, audio data is handled under strict operational guidelines:
| Component | Policy & Retention |
|---|---|
| Destination & Purpose | Sent to Google Gemini Developer API strictly for audio-to-text transcription. |
| Data Retention & Deletion | The Service deletes its temporary file after processing. It also attempts to delete the file uploaded to the Gemini Files API. If that deletion fails, the Files API automatically deletes the uploaded file after 48 hours. |
| No Conversation Storage | The transcription request sets store=false to opt out of Interactions API conversation storage. This does not control Files API retention or abuse-monitoring logs. |
| No Model Training | The voice feature requires an operational confirmation that it uses Paid Services. Under Google's Paid Services terms, inputs and outputs are not used to improve Google's products. |
* An explicit consent dialog is shown before recording begins. The microphone is never activated without your confirmation.
Even with Paid Services, Google may retain API inputs and outputs for up to 55 days for abuse monitoring and related purposes. The 48-hour Files API limit applies to uploaded files, not to every copy or log of the data.
Sources: Gemini Files API, abuse monitoring
5. Personally Identifiable Information (PII) Redaction & Masking
To prevent accidental exposure of sensitive information, our backend incorporates automatic privacy guards:
- Pre-storage Masking: Before submitted text is stored in session events or passed to language models, pattern-based filters replace email addresses, Japanese phone numbers, postal codes, and certain credit card number formats with placeholders. Names, employers, full addresses, and unusual formats may not be detected; please avoid entering them in your conversation.
- Log Minimization: The Service configures application logs and traces not to capture conversation or transcription text. Google's abuse-monitoring logs follow the conditions described above.
6. Session and History Management
- Guest Sessions (Signed-out): A temporary session ID is used for the conversation. After a page reload, the conversation cannot be resumed from the page and does not appear in My Page history. Closing the browser or starting a new conversation does not delete server-side session data. The guest identification cookie lasts 30 days by default.
- Signed-in Sessions (Google Sign-In): Only when you select “Save this conversation” is the masked conversation history saved to Google Cloud Firestore and linked to your account for My Page. Access requires authentication as that account. Unsaved conversations are not added to Firestore history.
- Cookie Usage: First-party cookies support authentication and guest identification. The guest identification cookie is also used for rate limiting. We do not use third-party advertising cookies.
Data Retention & Deletion Policy:
・Saved Consultation History (Firestore): As a valuable personal reflection log explicitly saved by you (user asset), saved history is safely retained as long as you do not request deletion (it will never be automatically deleted by the system).
・AI Inference Session (Agent Runtime): Session data maintains conversational context. The session and its conversation events are automatically deleted after the expiration time set when the session is created or updated. If no expiration is specified, Google's default TTL of 365 days applies.
・Data Deletion: You can delete individual saved conversations from My Page. This removes the saved Firestore history and related resume data, but does not delete the Agent Runtime session. To request deletion including server-side session data, use the Contact form in Section 10. We will verify your identity and arrange deletion of the relevant data. Agent Runtime sessions created while using the guest service before signing in are not linked to a Google Account and cannot be located from account details alone.
7. Security Measures
We employ multi-layered technical and organizational safeguards:
- Encryption: All network communication is enforced over TLS/HTTPS with modern cipher suites.
- Access Control: Google Cloud IAM principles of least privilege and strict service account isolation.
- Boundary Defenses: Request body size limits, IP and session-based rate limiters, and prompt injection detection guards. The default limit is 16 KiB for regular chat requests and 10 MiB for audio uploads.
8. Third-Party Services and Data Processors
We do not sell or rent personal information. Infrastructure hosting and AI computation are provided by the following vetted sub-processor:
- Google Cloud Platform / Google LLC: Serverless compute (Cloud Run), user authentication (Firebase Authentication), and AI APIs (Vertex AI / Gemini API).
9. Changes to This Privacy Policy
We may revise this Privacy Policy periodically to reflect service enhancements, legal requirements, or security updates. Updates become effective immediately upon posting to this page.
10. Contact Us
For questions or feedback regarding this Privacy Policy or your data, please contact us:
- Service: Arigato AI
- Contact: English contact form
About the Developer (Heysho)
ArigatoAI is an AI project planned and developed by Heysho, a creator working across digital marketing, SEO, AI search / GEO, data analysis, and workflow automation, while building software products with Python and generative AI.
* This project was developed as an open-source entry for the Google Cloud AI Hackathon.