Privacy & Safety

How we handle your family's data

No legalese. A clear explanation of where your data goes, what parents can and cannot see, how we handle crises, and how to delete everything.

The short version

Hearth is an AI wellness companion — not a therapist, not a medical device.
Your conversations are private to your account. Parents see summaries, never transcripts.
Data is sent to OpenAI for generation. It is not used to train models.
Crisis-level language triggers immediate safety resources and parent notification for minors.
You can export or permanently delete all your data at any time.

Where your data lives

•Conversations are stored in a PostgreSQL database hosted on secure infrastructure.
•Messages are sent to OpenAI's API for response generation. OpenAI does not use your data to train models (per their API data usage policy).
•No data is sold, shared with advertisers, or used for marketing purposes.

What 'private' means here

•Each account has its own isolated conversation history.
•Parents never see raw transcripts of their child's conversations.
•Parents receive summary-level insights: mood trends, session themes, and strengths — not what was said.
•Teens can enable teen-privacy mode, which further restricts what parents see to safety alerts only.

What parents can see

•Mood trends (e.g. 'anxious 3 times this week').
•Session themes (e.g. 'school', 'friendships').
•Strengths noticed (e.g. 'showed resilience').
•Usage stats: check-ins, message count, session minutes.
•Safety alerts if crisis-level language is detected.

Crisis detection & alerts

•Every message is screened for crisis-level language using keyword detection and OpenAI's moderation API.
•When detected: Hearth pauses, names what it noticed, and surfaces crisis resources for your region (988, Samaritans, local equivalents).
•For minor accounts (kid/teen): the parent is notified by email that a safety event occurred. The transcript stays private.
•Elevated-level alerts are throttled (max once per 4 hours) to avoid flooding. Crisis-level alerts always send immediately.

Child accounts & consent

•Children under 13 cannot create accounts on their own. A parent creates the child's space from the Family dashboard.
•The child receives an invite code — no email required for the child.
•The child sets their own username and password. The parent never has access to the child's login credentials.
•Teens (13–17) get age-appropriate language and safety. Parents can add teen accounts the same way.
•Hearth does not collect personal data from children beyond what the parent provides (name, age) and what the child shares in conversation.

Your data rights

•You can download all your data at any time from Profile → Your Data (JSON export).
•You can permanently delete your account and all associated data from Profile → Your Data → Delete.
•Deleting a parent account also deletes all linked child accounts and their data.
•There is no retention period after deletion — data is removed from the database immediately.

Our safety methodology

Hearth uses a two-layer safety system on every message:

Keyword layer: Pattern-matched against crisis-related phrases in 13+ regional variants (US, UK, CA, AU, NZ, IE, DE, FR, ES, PT, IL, IN, JP).
ML moderation: OpenAI's moderation API scores each message for self-harm, violence, and other categories.
Escalation: When either layer fires, Hearth immediately pauses, names what it noticed, and surfaces the right crisis number for the user's region.
Parent notification: For minor accounts, the parent receives an email alert. The child's transcript remains private.

Hearth prioritizes safety over engagement. If you need a real person, Hearth will tell you so — clearly and warmly.

If you are in immediate danger,please call emergency services (911 US, 999 UK, 112 EU) or reach the 988 Suicide & Crisis Lifeline (call or text 988). Hearth is not an emergency service and should not be relied upon in life-threatening situations.

Questions about privacy? [email protected]

Create your account