Evaluating AI Safety for Mental Health: Best Practices for Chatbots, Crisis Detection & Benchmarks

Описание: How do you evaluate whether an AI chatbot is actually safe for mental health? In this expert webinar hosted by the JED Foundation, leading clinicians, researchers, and AI developers share how they're evaluating and building safer chatbots — from open-source benchmark frameworks to real-time crisis protocols and human-in-the-loop review systems.

▶ WHAT YOU'LL LEARN
How AI chatbots are evaluated for mental health safety
Best practices for suicide risk detection and crisis intervention in AI systems
Why human-in-the-loop processes are the #1 predictor of AI safety
How open-source vs. closed-source models differ in child safety performance
What developers, clinicians, and institutions must align on before deploying AI tools

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
⏱ CHAPTERS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

00:00 Introduction — Laura Ericson, JED Foundation
04:00 Kora: Measuring AI Safety for Children — Stephie Hemlin
12:15 VERA-MH: Evaluating Chatbot Safety for Mental Health — Kate Bentley
22:40 Flourish Science: Building Evidence-based, Safe, and Effective AI for Mental Health — Xuan Zhao
36:30 Practical AI Safety Frameworks for Clinical Teams — David Cooper
46:30 Q&A: Frontier Models, Consensus, Parasocial Relationships & More

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🎙 SPEAKERS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🔹 Laura Ericson, MD — Chief Medical Officer, JED Foundation | Psychiatrist
🔹 Stephie Hemlin — Co-founder, Kora (child AI safety benchmark)
🔹 Kate Bentley, PhD — Faculty, Harvard Medical School | Senior Clinical Safety Officer, Spring Health | Creator of VERA-MH
🔹 Xuan Zhao, PhD — CEO & Co-founder, Flourish Science | Behavioral Scientist, Stanford | Creator of Sunnie AI | Author, Nature Reviews Psychology
🔹 David Cooper, PsyD — Clinical Psychologist | Chair, APA Mental Health Technology Advisory Committee | Associate Director, Digital Therapeutics, Otsuka

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔬 TOOLS & FRAMEWORKS MENTIONED
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Kora — Open-source child AI safety benchmark (25 risk categories, 32 models rated)
VERA-MH — Open-source AI mental health evaluation framework by Spring Health (GitHub)
Flourish / Sunnie AI — AI mental wellness app; scored 86/100 on VERA-MH
Columbia Suicide Severity Rating Scale (C-SSRS)
PHQ-9
Now Matters Now — crisis coping strategies resource
APA Ethics Framework crosswalk for AI (Tiffany Chenneville)
Working Alliance Inventory

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔗 RESOURCES & LINKS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

JED Foundation: https://jedfoundation.org
Kora benchmark: https://korabench.ai
Flourish Science: https://myflourish.ai
VERA-MH on GitHub: https://github.com/SpringCare/VERA-MH
Nature Reviews Psychology commentary by Xuan Zhao: https://www.nature.com/articles/s4415...
Now Matters Now: https://nowmattersnow.org

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

If you're building AI tools for mental health, evaluating vendors, or working in campus mental health, behavioral health, or clinical psychology — this conversation is essential viewing.

#AIsafety #MentalHealthAI #ChatbotSafety #AIethics #DigitalMentalHealth #SuicidePrevention #ResponsibleAI #VERAMH #Kora #FlourishScience #SpringHealth #JEDFoundation #ClinicalAI #ConversationalAI #AIchatbot #LLMsafety #BehavioralHealth #MentalHealthTech #AIinHealthcare #YouthMentalHealth