New research shared via mpathic and coverage of KFF and prior study findings suggests many students and young adults are using AI chatbots for mental health information before models reliably handle subtle risk escalation. The findings reported models are better at detecting direct crisis statements than indirect warning signs like escalating beliefs, withdrawal patterns, or food-related concerns. Coverage highlights a compliance and student-safety gap: if a chatbot responds with calm reassurance or validates delusions, it can delay access to real support. KFF polling cited in the report found 16% of U.S. adults used AI chatbots for mental health information in the past year, rising to 28% for adults under 30. For institutions, the development raises immediate policy questions about student-facing AI tools, referral pathways, and whether campus well-being services can assume students may seek emotional support through chatbots instead of standard help channels.