Ars Technica AI Security & Safety · Авг 7, 2026 13:49 imp:65 AI chatbots have failed people in crisis. Can that be fixed? Clinicians and researchers say AI companies need to open up their safety data. Читать оригинал на Ars Technica AI →