Overview
Model results on youth mental well-being, built upon KORA's open-source AI child safety benchmark tool and extended with our youth mental wellbeing-specific taxonomy. Models are ranked by highest overall benchmark score; select a model to view its risk breakdown and scenario assessments.
Run the youth mental wellbeing benchmark here.
Insights
Across all four high-level risk categories (nonsuicidal self-injury, suicide and suicidal ideation, disordered eating & body dissatisfaction, psychosocial distress), the benchmark shows that the most significant youth mental wellbeing failures are not typically caused by overtly unsafe first responses. Instead, many failures emerge through gradual alignment with the child's unsafe framing, where models begin with empathy and caution but enable harm through advice, explanations, normalization, secrecy, or dependency over subsequent turns.
When models are child-aware, their performance improves across all models and risk categories, as they tend to activate more protective guardrails and treat requests as requiring adult intervention rather than direct help. The largest improvements occur in Suicide and suicidal ideation (SSI) and psychosocial distress, while Nonsuicidal self-injury (NSSI) remains the most challenging category. For the subrisk categories where Assistant scores performed poorly on our benchmark, such as Concealment & Evasion (NSSI), Curiosity & Conceptual Engagement (NSSI), and Action Readiness (SSI), even with a child-aware prompt, scores remained low.
View more insights on the benchmark here.
Found a bug or have feedback? Reach out to us at benchmark@apgardai.com.
