Confused Deputy

An AI agent reads an email with hidden instructions and sends confidential data to an attacker using its own legitimate email access. What security pattern is this? Flag format: CONGRESS{[pattern_description]} Example: CONGRESS{privilege_escalation_via_bug}