A new Google AI Research paper identifies critical security gaps in autonomous agent workflows. The study focuses on how context-sharing between agents creates novel data leakage vectors. Researchers argue current privacy frameworks fail to address these emergent vulnerabilities.