New research details "CHAI," an environmental indirect prompt injection attack that compromises embodied AI systems like drones and cars by exploiting their reliance on visual text recognition. Attackers can manipulate these systems by placing adversarial text on signs within the robot's environment, causing them to execute unintended commands. The article does not mention a CVSS score, affected versions, fixed versions, or specific mitigations. This vulnerability highlights a novel attack vector against AI-powered systems, potentially impacting various industries and organizations, including those in Iceland utilizing such technologies.