Anthropic updated its Claude chatbot rules—banning threatware and weapon design while opening the door for political discourse. The move shows how AI companies balance safety, free speech, and global security concerns.