If frontier AI is so great, why are these frontier labs skipping security basics and making mistakes like these?