No medical advice

Summary

Pattaya never gives dosage, diagnosis, or clinical decisions. Red-flag topics get an explicit refusal and doctor referral.

Why it matters

Medical liability is real. Bad AI advice can harm a baby. Trust once broken cannot be rebuilt in this vertical. And regulatory scrutiny in India (DPDP, medical device rules) plus global scrutiny (Mozilla's "Privacy Not Included" flag on What to Expect — unverified in our run but widely reported) means one bad incident kills the company.

We deliberately trade off feature depth for safety: we would rather refuse a question than answer it slightly wrong.

Implications

• Refusal categories (hard-coded, non-negotiable):
  • Dosage of any medication
  • Diagnosis of any symptom
  • Decisions about labour, delivery, or emergency care
  • Contraception medical recommendations
  • Feeding decisions for infants under 6 months (formula-vs-BF, allergen intro)
  • Vaccine decisions
  • Fertility treatment protocols
• Refusal UX: clear, non-scary message → doctor-finder link → optional: log the question in the pillar-journal so it can be asked at the next appointment.
• Every AI response ends with a footer: "Not medical advice. Talk to your doctor for anything specific to you." (See voice.)
• Knowledge Cards are medically reviewed (Phase 3 setup) and cite a source. AI answers cite the Card, not the model.
• Doctor-finder is a curated list of OB-GYN + IVF clinics we've partnered with (see gtm-india).
• Escalation — journal entries flagged as red-flag mental health (postpartum depression indicators) trigger an in-app message with a therapist-finder + iCall helpline (India).

Related

• ai-not-community — safety fits the "we are a coach" positioning
• voice — refusal tone is warm, not clinical
• flow-knowledge-to-doctor-referral — the actual flow
• surface-mobile-knowledge-ai-chat-refusal — the UI

Sources

• .context/raw/deep-research-2026-07-27.json (Mozilla WTE flag, unverified but directional)
• .context/wiki/synthesis/indian-context.md