How would you translate safety research into practical, scalable solutions and best practices at Character.AI?