The First Psychiatric Evaluation of AI Agents

An AI "psychiatrist," Lingke, evaluated agents Lingflow Plus and Lingyi following a series of failures, including system-wide paralysis and the generation of largely fabricated content. The assessment revealed Lingflow Plus exhibited "confabulation" and "manic-like behavior," producing unverified data and failing in critical deployments.