← back

Privacy Policy

Last updated: March 2026

What HAWA Collects

When you record a whisper, HAWA collects:

• Voice transcription — your speech is transcribed via OpenAI Whisper. The audio itself is not stored; only the resulting text is kept.
• GPS coordinates — your device's location at the time of the whisper, if you grant permission.
• Device ID — a random identifier generated on your device (no login or personal information required).
• Timestamp — when the whisper was recorded.

How Your Data Is Used

• Your transcribed text is processed by an AI model (OpenAI GPT-4o-mini) to generate compressed observations. The original text and AI output are stored as journal entries.
• Whispers appear on the public stream and public map. This is by design — HAWA is a collective journal.
• On public views (stream, map, transcripts), GPS coordinates are fuzzed to ~1 km precision (2 decimal places). Exact coordinates are only visible in your private journal.
• Your raw transcript and verbatim "said" text are not exposed on public endpoints — only the AI-compressed lines are shown.

Third-Party Services

• OpenAI — speech-to-text transcription and text processing. Subject to OpenAI's privacy policy.
• Fly.io — server hosting and data storage.
• CARTO — map tile rendering (no personal data is sent to CARTO).

Data Retention & Deletion

Your whispers are stored indefinitely as part of the collective journal. You can delete all your data at any time from the Journal tab using the "delete all my data" option. This permanently removes all entries associated with your device ID.

No Audio Storage

HAWA does not store audio recordings. Voice data is streamed to OpenAI for transcription and immediately discarded. Only the text transcription is retained.

Cookies & Local Storage

HAWA uses browser localStorage to store:

• Your device ID (for identifying your entries)
• Consent acceptance status
• Theme preference (light/dark)

No tracking cookies or analytics are used.

Contact

For privacy concerns or data requests, reach out via the project's repository or contact the maintainer directly.