Tombiby Chamomile Lab
Updates ↗
Menu

PRIVACY AT TOMBI

Your assistant.
Not everyone’s recorder.

Tombi’s redaction design aims to keep other people’s speech out of your retained recordings, transcripts, and summaries.

Read the pre-launch privacy policy draft ↓

PRIVACY BY DESIGN

Built around you.
Respectful of others.

Useful personal AI should understand your life without making everyone around you part of your personal record.

01

Purposeful personal context

Information should support the assistance you choose, with clear boundaries around how it is used.

02

Respect for other people

Our redaction goal is to exclude other people’s private words from retained personal context.

03

Choice stays with you

Choose your features and sharing preferences. Your conversations and memos are not part of family sharing.

WHY REMOVING AUDIO ISN’T ENOUGH

Words can outlive
the recording.

Raw audio

A recording can include another person’s voice and private information, even when the device is worn by you.

Transcripts

Deleting the sound file does not delete a written copy of what someone said. Speaker labels alone are not redaction.

Summaries

A summary can preserve someone’s private details without quoting them. Redaction and deletion need to cover derived records too.

PRE-LAUNCH PRIVACY POLICY · DRAFT

What you should know
before you choose Tombi.

1. What information supports the service?

Account and consent records; preferences, goals, confirmed reminders, connected calendar information, wearer-attributed requests and responses; permitted location and sensor data; device status; and optional safety, support, or research records. Information about health, routines, and relationships may be sensitive.

2. How do we approach conversation privacy?

Tombi is designed to exclude other people’s speech from retained personal assistance records. Release-specific notices will explain the information used by each feature, the choices available to you, and applicable retention terms.

3. Is everything processed on my device?

No such promise is made. A consented test build may keep personal-context records in the cloud with authorized staff access. The planned standard service keeps the assistance record primarily on the phone, while some speech and AI functions may still use cloud processors. Your release notice must explain which arrangement applies.

4. What do we use the information for?

Deliver requested assistance, manage confirmed reminders, answer questions, apply preferences, connect devices, support the account, and maintain security and reliability. Research, cross-user learning, and other secondary uses require separately disclosed authorization where required. Accepting the service does not by itself authorize those uses.

5. Who can see my information?

Service providers may process information needed for enabled functions. Authorized staff may access information for disclosed test, support, security, or research purposes. Trusted people see only permitted categories; buying the device or helping with setup does not automatically grant access. Conversations are excluded from the proposed trusted-person view. The draft prohibits selling personal information and using it for cross-context behavioral advertising; provider arrangements must support that commitment.

6. How long is information kept?

The feature plan proposes one year of personal conversation history. The policy draft still requires a final schedule for transcripts, summaries, audio buffers, research data, location, logs, backups, and deletion from disconnected devices. These are different data types, not one blanket retention period. No final deletion deadline or zero-retention vendor promise is asserted on this preview.

7. What choices and rights will I have?

The proposed controls cover access, export, correction, deletion, connected sources, and individual sharing permissions. Optional research has its own consent and withdrawal process. Removing data needed for a feature may disable that feature. Request channels and release-specific app controls must be published before use.

8. How is research different?

Research may involve additional raw signals or audio only under a separate, specific authorization with collection and deletion dates. A testing window is not permission to retain information indefinitely. Pseudonymous data can still identify someone and must be treated accordingly.

9. What about security, health information, and children?

The service is intended for adults. Health-related information requires appropriate sensitive-data handling; consumer use alone does not establish HIPAA coverage. Security controls and incident procedures must match the actual release. This preview claims neither perfect security nor an independent certification.

10. What will be finalized before launch?

The responsible legal entity and privacy contact; actual processors and processing locations; the final retention and deletion schedule; verified security controls; applicable state and consumer-health notices; permission flows; and the effective date. A change in policy must not silently authorize a new sensitive use.