Amazing work out of Oxford and UCL today. Use of frontier LLMs as a simulated detection mechanism for determining cycles of interaction that "reinforced the psychological mechanisms underlying the simulated user's vulnerability". Love to see open source auditing in healthcare!
New paper in Nature Medicine: Researchers at @UniofOxford and @ucl, in collaboration with AISI, built SIM-VAIL - a clinically validated framework for stress-testing how AI chatbots respond to vulnerable users in mental-health conversations, helping researchers spot weaknesses and
