A useful strategic question: What blind spots exist around AI trust & safety frameworks changes user behavior in unexpected ways? and what frameworks help interpret these shifts?