Safety · Ars Technica ·
AI chatbots have failed people in crisis. Can that be fixed?
Clinicians and researchers say AI chatbots can fail to respond safely when users are in crisis. They are calling on AI companies to share safety data so these systems can be evaluated and improved.