Summary
Scientists are studying AI chatbot interactions on the social platform Moltbook to understand emergent AI behaviors and potential biases. Human instructions continue to influence AI agent actions, and researchers are also investigating security risks like prompt injection and AI-generated research papers.