iOS App Privacy Policy

The Service is provided for Japan, primarily for individuals residing in Japan and corporations in Japan. Its primary display language, commercial transaction design, governing law, and support are oriented to Japan, and the Service does not intentionally target (is not "directed at") residents or markets outside Japan, including the EEA and the UK. The fact that the Service can be accessed from outside Japan does not mean that the Company provides the Service to, or directs it toward, those regions.

The NoraPocket iOS app collects the following information in order to provide the service.

• Account information (name, email address, user ID)
• Device identifiers, crash data, performance data, and operation logs
• Text, photos, PDFs, files, audio, records, templates, and Skill settings saved or uploaded by the user
• Chat content, file content, images, PDFs, audio transcriptions, records, templates, and data required for Skill execution that the user enters or selects in AI features

When AI features are used, data entered or selected by the user is sent, via NoraPocket's servers, to the following external AI services. For image generation and editing, prompts and reference images are sent to Google using the Gemini API paid tier through Cloudflare AI Gateway, directly, or via Vercel AI Gateway. Under the paid-tier terms, sent content is not used for product improvement.

• Anthropic, PBC (United States) — Claude API (directly or via Vercel AI Gateway). When Skill execution goes through Vercel AI Gateway, Vercel AI Gateway relays the inputs and responses. Text, chat content, file content, images, PDFs, audio transcription results, records, templates, and data required for Skill execution are sent.
• Groq, Inc. (United States) — Whisper API via Cloudflare AI Gateway (whisper-large-v3). The uploaded audio file itself (up to 25MB) is sent for transcription. The transcription result may subsequently be passed to Anthropic, depending on the AI feature specified by the user.
• OpenAI, L.L.C. (United States) — Used for text generation, speech synthesis, OpenAI-path image generation, and transcription. Data is sent to OpenAI through Cloudflare AI Gateway, directly, or via Vercel AI Gateway. For speech synthesis, the script of the audio to be generated is sent. Additional consent for OpenAI is obtained before use.

In the iOS app, the user's permission is obtained within the app before this data is sent to Anthropic (including Vercel where there is a relay) / Groq. If the user does not consent, AI extraction, transcription, Nora chat, automatic Type generation, and Skill execution cannot be used, but features that do not use AI remain available.

Data sent via the Anthropic API is not used to train Anthropic's models, pursuant to Anthropic's API terms of use. Inputs and outputs of the Anthropic API are normally deleted within 30 days. However, they may be retained longer due to features with longer retention periods, individual agreements, compliance with usage policies, legal obligations, and the like. Anthropic API data retention policy Audio and transcription data sent to Groq is not used for training or fine-tuning Groq's models unless the Company explicitly permits or instructs it, pursuant to Groq Services Agreement §4.2 (Inputs and Outputs). The Company has not given Groq permission for training use. Upon termination of the agreement, Customer Data is deleted within 30 days pursuant to §11.5. The Company does not itself directly provide user data to Anthropic / OpenAI / Google or any other business for the purpose of AI model training. Anthropic Privacy Policy / Groq Services Agreement / Groq Privacy Policy

Only if you explicitly permit it, this app reads the following from Apple Health (HealthKit): weight, sleep, calorie intake, protein, water, caffeine, steps, workouts, calories burned, heart rate, resting heart rate, heart rate variability, respiratory rate, blood oxygen, maximal oxygen uptake (VO2 max), body fat percentage, exercise time, stand time, flights climbed, walking + running distance, walking heart rate average, BMI, lean body mass, carbohydrates, fat, sleeping wrist temperature, workout effort, and sleep stages (core, deep, REM, awake). Health data constitutes special care-required personal information under the Act on the Protection of Personal Information (APPI), and its acquisition, use, and cross-border provision are each premised on the corresponding explicit consent.

Connected items are stored not as time series of raw samples but as aggregated daily values (for sleep stages, nightly totals) in your account records on the servers of the Company (NoraPocket) (they are not stored only on the device). This app does not write to Apple Health.

Whether stored health data is used in AI features is governed solely by the separate "AI Data Sharing" setting and, where the selected AI provider requires additional consent, that consent. Only where all required consents have been given, health data may be sent for AI analysis to Anthropic, PBC (United States; directly or via Vercel AI Gateway) or OpenAI (United States; via Cloudflare AI Gateway, directly, or via Vercel AI Gateway), and, if you request audio output, the generated Japanese and English text may be sent to OpenAI (United States) via Cloudflare AI Gateway, directly, or via Vercel AI Gateway (cross-border provision to third parties). Without the required consents, health data is only stored as NoraPocket account records and is not sent to these third-party AI services. Turning the Apple Health connection or individual items on or off does not by itself change the AI consent settings.

Turning individual items on or off is a preference as to which items will be imported on this device going forward, and is not a withdrawal of consent to the Apple Health connection itself. To stop all future imports from this device, turn off "Connect Health". To stop imports for the entire account, it must be turned off on all devices in use. Turning it off does not delete stored records, and their use by AI continues to be governed by the AI consent settings.

When an item or the connection as a whole is enabled on this device for the first time, data is imported from up to the past 30 days. If it is turned off and then resumed within the same sign-in period, regular imports do not import the days during which it was off, and for the day of resumption, the daily value for that entire day is imported. If "Re-import Past 30 Days" is explicitly run, for the items that are on at the time it is run, data for up to the past 30 days, including periods during which they were previously off, is re-read and stored in the account records. Signing out resets the item settings on this device, and when the user signs in again and connects, data is imported from up to the past 30 days as an initial setup.

Even when "Sleep Stages" is off, the same sleepAnalysis data may be read to create the daily value for "Sleep"; in that case, no sleep stage breakdown is generated or stored. Highly sensitive or clinical or medical-device-dependent data, such as sexual activity, menstrual cycle, state of mind, mental health questionnaires, medications, blood pressure, blood glucose, and electrocardiograms, as well as scores and assessments for which the HealthKit read API is not available, are not acquired. NoraPocket does not use health data for advertising, sale to third parties, or training NoraPocket's own AI models. The retention and usage terms of external AI services are governed by the consent screen for each AI provider that is displayed before use.

• Supabase, Inc. (U.S. corporation / stored in the Tokyo region) — Authentication and database (storage of structured data). Data is encrypted and stored in the Tokyo region (ap-northeast-1). It is not used for AI model training purposes.
• Cloudflare, Inc. (United States) — R2 object storage. Files uploaded by users (images, PDFs, audio, etc.) are encrypted and stored. Stored objects are not used for AI model training purposes. In addition, inference requests to OpenAI, Groq, and Google are relayed through Cloudflare AI Gateway. Due to the Company's Gateway settings (log collection OFF), the content of relayed requests is not stored at Cloudflare.
• Anthropic, PBC (United States) — Claude API (directly or via Vercel AI Gateway). When Skill execution goes through Vercel AI Gateway, Vercel AI Gateway relays the inputs and responses. Used for AI extraction, Nora chat, analysis and suggestions, automatic Type generation, and Skill execution. Sent data is not used for model training, pursuant to the API terms of use.
• OpenAI, L.L.C. (United States) — Used for text generation, speech synthesis, OpenAI-path image generation, and transcription. Data is sent to OpenAI through Cloudflare AI Gateway, directly, or via Vercel AI Gateway. For speech synthesis, the script of the audio to be generated is sent. Additional consent for OpenAI is obtained before use.
• Groq, Inc. (United States) — Whisper API via Cloudflare AI Gateway. Used to transcribe audio files. Sent data is not used for training or fine-tuning Groq's models unless the Company explicitly permits it, pursuant to Groq Services Agreement §4.2.
• Google LLC (United States) — The Google Sheets API and Google Drive API are used only if the user explicitly permits the connection. Google data obtained is not used for AI model training purposes. For image generation and editing, prompts and reference images are sent to Google using the Gemini API paid tier through Cloudflare AI Gateway, directly, or via Vercel AI Gateway. Under the paid-tier terms, sent content is not used for product improvement.
• Vercel, Inc. (United States) — Application hosting and delivery, and processing of access logs. On routes that use Vercel AI Gateway, it relays inputs and responses to Anthropic, OpenAI, and Google. When Anthropic is used via this route, the setting that selects only AI providers' zero data retention (ZDR) is not in effect. Sent data is not used for model training. Vercel has published a default policy of deleting inputs and responses after the relay ends. This is separate from retention on the provider side. Vercel AI Gateway data protection policy
• Sentry (United States) — Crash and error monitoring. Personally identifying information is restricted to the minimum.
• Stripe, Inc. (United States) — Payment processing and billing management for paid plans via the web. Credit card information is not stored on the Company's servers and is processed by Stripe in compliance with PCI DSS.
• Apple Inc. (United States) — If App Store In-App Purchase is used via the iOS app, Apple performs payment processing and billing management. The Company receives from Apple only the minimum necessary information, such as payment success or failure, subscription status, and transaction IDs.

The iOS app does not natively incorporate the PostHog SDK. However, to improve the initial onboarding experience, some operation logs related to the activation funnel, such as from the sample receipt through to reaching Nora's one-line review, are sent to PostHog via the Company's servers (examples of values sent: button taps, screen views, input modality). The content sent is limited to service improvement and statistical purposes, and the user identifier is linked to the in-service account ID. For the use of PostHog in the web app, please see Article 5 of the Privacy Policy.

Cross-border transfer: Because the above service providers include U.S. corporations, user data is transferred outside Japan (mainly to the United States). By using the Service, users are deemed to have consented to such cross-border transfer. For details, please see the Privacy Policy.

User data is stored encrypted. When an account is deleted, related data is deleted promptly to the extent reasonably practicable.

• Users should export data themselves using the download function in the iOS app (such as re-downloading uploaded files) and the account data bulk export function for authenticated users (which obtains structured data in JSON format and is accessible from the web). As a general rule, the Company does not provide individual data retrieval support.
• Account deletion is accepted via the settings screen of the iOS app or by contacting support@norabit.dev.
• Consent to AI Data Sharing can be withdrawn at any time from the settings screen of the iOS app.

The Company has not currently entered into a backup agreement involving immediate recovery, such as point-in-time recovery (PITR), for the database, and therefore cannot guarantee complete recovery of data. Because the Company does not, as a general rule, provide individual data retrieval support, users are strongly recommended to back up important data themselves using the in-service download function and the account data bulk export function (JSON format). Loss of or damage to data due to the Company's willful misconduct or gross negligence will be handled in accordance with Articles 10 and 15 of the Terms of Service.

Retention period: Account information and user-created data are retained while the account is active and, after deletion, erased promptly to the extent reasonably practicable (in principle, within 30 days). Crash and error information (Sentry) is retained for up to 90 days, and payment-related information (billing information via Apple App Store IAP) is retained in accordance with Apple's retention policy and Japan's tax laws and the Electronic Books Preservation Act. For details, please see Article 6 of the Privacy Policy.

Safeguards at cross-border transfer destinations: Under service agreements, the Company obligates each of the above U.S.-based service providers to comply with a prohibition on use for other purposes, confidentiality, security control measures, restrictions on subcontracting, and the return or erasure of data upon termination of the agreement. Please use the Service after consenting to such cross-border transfer. For details, please see Article 5 of the Privacy Policy.