This is delightful. As someone who uses AI/ML/MI/... for security, I find there is not nearly enough understanding for how attackers can subvert decision systems in practice.

Keep up the good work!