As companies automate e-mail and workflow decisions with AI, attackers are exploiting those same systems, launching AI-to-AI phishing attacks that bypass human oversight.