Responsible AI & Safety
Every design decision starts with the safety of the people who need this most.
Our AI Ethics
- Pattern analysis, not diagnosis. Receipts identifies communication patterns — it does not diagnose conditions, assign blame, or make legal determinations.
- Human interpretation first. AI outputs are designed to support human professionals — therapists, lawyers, advocates — not replace them.
- No training on your data. Your conversations are never used to train AI models. Analysis is performed via secure API calls and results are returned to you.
- Bias awareness. We actively work to account for cultural, linguistic, and contextual factors that may affect pattern detection.
- Transparency. We explain how patterns are detected, what confidence levels mean, and where limitations exist.
Safety by Design
Panic Button
Instantly exit the app and clear the screen. Designed for situations where safety depends on discretion.
App Disguise
Receipts can appear as Calculator, Notes, or Weather on your device. No visible trace of its real purpose.
No Sensitive Notifications
Push notifications never contain message content, pattern names, or contact information.
PIN & Biometric Lock
Additional authentication layer beyond your device lock screen.
Self-Pattern Detection
When the AI detects that the user may be exhibiting concerning patterns, it surfaces this gently with resources — not judgment.
Crisis Resources
If you or someone you know is in immediate danger, contact emergency services in your area.
- Australia: 1800RESPECT (1800 737 732) — 24/7 domestic violence and sexual assault helpline
- United States: National DV Hotline (1-800-799-7233) — 24/7 domestic violence support
- United Kingdom: National DV Helpline (0808 2000 247) — 24/7, run by Refuge
- Canada: Assaulted Women's Helpline (1-866-863-0511) — 24/7
- New Zealand: Women's Refuge (0800 733 843) — 24/7 safe house and support
- Ireland: Women's Aid (1800 341 900) — 24/7 national freephone helpline
In an emergency, always call your local emergency number (000, 911, 999, 111).
Warrant Canary
Sometimes governments ask companies for user data and legally forbid them from telling anyone about it. We can't say "we received a secret order" — but we can say "we haven't received any secret orders" and stop saying it if that changes.
We maintain a signed statement that we update at least every 30 days. As long as it's being updated, you know:
- We have NOT received any National Security Letters
- We have NOT received any secret court orders (FISA, etc.)
- We have NOT been subject to any gag orders
- We have NOT been forced to add backdoors or weaken security
- We have NOT handed over encryption keys
What to watch for:
- File stops updating: If the file isn't updated for more than 30 days, something may be wrong
- Statement removed: If specific statements disappear, that category may be compromised
- File disappears: If the canary goes away entirely, assume the worst
You're trusting us with sensitive information about your relationships. If we're ever forced to compromise that trust, we want you to know — even if we can't tell you directly.