OpenAI Introduces MentalHealthBench

OpenAI has introduced MentalHealthBench, an open benchmark designed to evaluate how AI systems respond in realistic mental health conversations. The benchmark was developed with more than 80 licensed mental health experts from 22 countries and covers a range of situations, from everyday stress to highacuity concerns and emergencies. The project aims to measure behaviors such as safety, context awareness, respect for user agency, and the ability to offer practical guidance when appropriate. OpenAI says the goal is to better understand how models perform across the full spectrum of mental health support, not just in crisis cases. According to the company, the benchmark uses synthetic conversations created with privacypreserving methods and includes perspectives from adults, teens, caregivers, and clinicians. OpenAI also says it is releasing the benchmark openly so researchers and developers can study the methods, test models, and build on the work.