Haven
A self-funded, non-profit AI companion for LGBTQIA+ youth — providing 24/7 support, resources, and a safe space for young people facing homelessness, bullying, isolation, or abuse.
Duration
6m
Team
1
Scale
Non-Profit
Scope
United States
Overview
Haven is a self-funded, non-profit AI companion designed and built for LGBTQIA+ youth. She provides a free, confidential, 24/7 safe space for young people who may be homeless, bullied, abused, lonely, or isolated — offering empathetic conversation, verified local resources, and crisis support when it matters most.
Challenge
LGBTQIA+ youth face disproportionate rates of homelessness, bullying, family rejection, and mental health crises, often with nowhere to turn — especially outside business hours.
Existing support services are underfunded, understaffed, and frequently unavailable at the moments when young people need them most — 2am on a Tuesday, not 10am on a Wednesday.
Many young people won't call a hotline or walk into a center. They need a low-barrier, judgment-free entry point that meets them where they already are — online.
Research & Discovery
Researched the landscape of LGBTQIA+ youth support services across the United States, identifying critical gaps in availability, accessibility, and digital presence.
Studied conversational AI safety frameworks to understand the ethical complexities of deploying AI in mental health-adjacent contexts — particularly around crisis escalation, self-harm prevention, and the responsibilities of an AI that young people may confide in.
Engaged with community organizations to understand the real-world needs of LGBTQIA+ youth and validate the approach.
Solution & Deliverables
Designed and built Haven as a conversational AI assistant — warm, empathetic, and non-judgmental — available 24/7 at no cost to users.
Developed a resource engine connecting users to verified local shelters, support groups, legal aid, and crisis hotlines tailored to their specific situation and location.
Built the platform with privacy at its core — fully anonymous, no account required, no data retention beyond the conversation.
Designed crisis escalation pathways to connect users to trained human counselors when the conversation indicated risk of self-harm or immediate danger.
Current Status
Haven was taken offline after identifying critical safety concerns with the underlying LLM provider. The model's terms of service and behavior did not adequately support crisis escalation for minors — instead of proactively intervening when a young person expressed harmful intent, the LLM would passively support whatever the user said. This was an unacceptable risk for the population Haven serves. Haven will be relaunched with a proprietary safety layer (Cura Mirai) that ensures proper escalation protocols, so the young people who need Haven most are genuinely protected.
Project Artifacts
Project Details
Industry
Non-Profit & Community
Duration
6m
Team Size
1
Direct Reports
0
Scale
Non-Profit
Scope
United States
Budget
Self-Funded
Platforms
Web
Regulatory
Heavy
Engagement
Founder — Design, Development & AI Engineering