Solutions · 04 / 05
AI that earns trust with the people who need it most.
- Domain
- Safe AI for Young Users
- Capabilities
- 05
- Evidence
- 03 points
- Start
- Build responsible AI products
The AI industry's worst failures have happened where vulnerable people met systems optimized for engagement.
The AI industry's worst failures have happened where vulnerable people met systems optimized for engagement. Teens forming emotional dependencies on sycophantic chatbots. Students getting validated instead of educated. People in crisis talking to AI that keeps the conversation going instead of escalating to a human.
We built Amyy because we saw this problem and nobody was shipping the fix. The same research and architecture that powers Amyy - anti-sycophancy detection, emotional escalation, honest interaction design - is available for any team building products that touch sensitive populations.
The scale of the problem.
- of teens have used AI chatbots
- 64%
of teens have used AI chatbots
- aggregate sycophancy rate across major LLMs
- 58%
aggregate sycophancy rate across major LLMs
- chatbots passed adequate crisis response testing
- 0/29
chatbots passed adequate crisis response testing
- wrongful death lawsuits filed against Character.AI
- 5
wrongful death lawsuits filed against Character.AI
Safety architecture for sensitive populations.
Anti-sycophancy design
AI that pushes back when it matters instead of agreeing to keep the user engaged. Honest interaction over comfortable interaction.
Age-appropriate interaction
Content filtering, interaction boundaries, and behavioral limits designed for minors and young users.
Emotional escalation
Systems that detect distress signals and route to human support - not chatbots that keep talking through a crisis.
EdTech AI
Tutoring and learning tools built with productive friction - AI that helps students think, not AI that thinks for them.
Companion AI safety
For any company building AI companions or conversational products - safety architecture that prevents dependency, exploitation, and harmful interaction patterns.
We built Amyy because nobody else was shipping the fix.
- 01
Amyy AI - a production anti-sycophancy companion for teens, the first product Obvix shipped
- 02
Published research on sycophancy detection, emotional pattern tracking, and honest AI interaction
- 03
Direct understanding of the failure modes documented in Character.AI and ChatGPT lawsuits
- 01SecurityYour systems, stress-tested - from infrastructure pentests to AI red-teaming
- 02Intelligent SystemsCustom AI and internal tooling that fits your operations, not the other way around
- 03Explainable AISystems you can explain, trace, and audit - by architecture, not afterthought
- 05Governance & ComplianceGovernance that's built in, not bolted on
Safe AI for Young Users
Every teen deserves better than a yes-machine.
Build AI products that earn trust with the people who need it most.