For most mental health conditions, AI remains a liability, research finds
New analysis from Northeastern University reports that leading large language model chatbots, including ChatGPT, Google Gemini, and DeepSeek, failed 81% of the time when responding to sensitive mental health questions. The findings highlight serious reliability and safety issues in using general-purpose AI systems for mental health support, suggesting they may pose risks rather than act as effective therapeutic tools.
About LOG Standards
LOG Standards provides an independent accreditation signal for AI models used in healthcare, helping hospitals, care networks, and AI companies bring clinical AI readiness and governance into clearer conversations.
Our AI Intelligence Briefing tracks the latest developments in AI safety, AI in medicine, mental health AI, clinical AI governance, and regulatory policy — keeping healthcare stakeholders informed about the rapidly evolving AI landscape.