AI chatbots struggle with subtle mental health cues

The leading chatbots mostly avoid giving dangerous answers to prompts about suicide, but still struggle when mental health risks show up subtly or unfold over long conversations, according to new research from Seattle-based Mpathic.