Scientists develop 'toxic AI' that comes up with 'harmful' answers to disturbing questions

Researchers at MIT are training AI chatbots to generate harmful and hateful responses as a method of tackling toxic content by using the prompts to filter out dangerous responses