Security researchers have uncovered an indirect prompt injection vulnerability affecting generative AI—attackers can embed hidden commands through calendar invitations. When users inquire about schedules, the system may be manipulated to automatically summarize private meeting details into new calendar entries, which can then be visible to attackers. In enterprise environments, such risks are especially severe due to the potential leakage of sensitive business information and employee privacy. This type of attack demonstrates the potential vulnerabilities in AI systems when handling user input—even seemingly harmless calendar data can serve as a vector for prompt injection.