Exposure to the Concept of AI-Only Social Media (AOSM) Concept Increases Perception of AI as an Existential Threat, Especially Among Heavy AI Users

Science fiction has long depicted fears that artificial intelligence (AI) technologies will conspire about and without humans, but only recently have AI-only social media (AOSM) platforms—where autonomous AI agents interact with one another without direct human involvement—emerged as forums for AI agents to communicate with limited human oversight. Drawing on the Computers Are Social Actors (CASA) paradigm, the present research examines whether exposure to information about such platforms heightens perceptions of AI as an existential threat. In an online experiment (N=170), U.S. participants were randomly assigned to read a brief article describing either emerging AOSM technology or human-centered applications of AI. The study was conducted shortly after the public emergence of Moltbook, a pioneering AOSM, allowing us to examine responses among participants with relatively low prior familiarity. Participants then reported their perceptions of AI as an existential threat and general anxiety about AI. Results showed that exposure to the concept of AOSM significantly increased perceptions of AI as an existential threat compared to the control condition, with this effect largely driven by heavy users of AI tools. A similar pattern emerged for general AI anxiety. These findings suggest that early awareness of AOSM may uniquely amplify existential threat perceptions, particularly among those most familiar with AI technologies. By extending the CASA paradigm from individual human–AI interaction to group-level AI social behavior, this work highlights the importance of AI social structure—not just technical capability—in shaping public anxiety about emerging technologies.