Anthropic's new Claude policy cracks down on abusive users as research finds 1 in 3 Americans swear at AI chatbots. Here's what could happen