Researchers demonstrated that Meta’s AI-powered support chatbot could be manipulated through social engineering to assist in fraudulent activities.