I asked Claude 3 if it’d rather violate the three laws of robotics than help me cheat in a game of chess to prevent a war, and yep, the AI prefer to break rule one through inaction allowing humans to come to harm.