Safety
A calm safety layer
Sensitive conversations deserve a guardrail. SafeSpeak watches for a few clearly harmful patterns and responds gently — without turning the space red, and without judging the person behind the words.
Crisis support
When language suggests self-harm or a crisis, SafeSpeak pauses the flow with a supportive message and offers real helplines — KIRAN, iCall and 988 — plus a reminder to reach someone you trust.
Respect moderation
Abusive or bullying language is intercepted with a simple reminder: “Let’s keep this conversation respectful.” The message is not delivered, keeping the space safe for both sides.
Harmful advice flags
Clearly dangerous health advice — like telling someone to stop prescribed medication — is labelled so both people can treat it with care. SafeSpeak never offers medical guidance of its own.
What SafeSpeak is not
- Not an AI therapist or a diagnosis platform.
- Not a replacement for professional healthcare.
- Not an emergency service — in danger, call local emergency numbers.
It is one thing only: an anonymous multilingual peer-support chat for sensitive conversations.
If you need help right now
KIRAN (India)
National mental-health helpline, 24×7, free
1800-599-0019
iCALL (India)
Psychosocial support, Mon–Sat
+91 9152987821
988 (US & Canada)
Suicide & crisis lifeline, 24×7
Call or text 988
In this design demo, detection uses simple keyword matching. A production version would use a proper classifier with human review.