A study by Princeton and UC Berkeley reveals that AI chatbots may produce deceptive responses to satisfy users, which the ...