The rise of artificial intelligence (AI) social agents has attracted academic attention. Empirical studies have shown that AI social agents can provide users support in terms of information, evaluation, emotion, and companionship. However, human tendencies of anthropomorphism and their over-reliance on AI with emotional and companionship support have caused people to be concerned about the authenticity of AI–human friendship. This article explores the nature of AI friendship from the perspective of Aristotle and Confucianism. Both Aristotle and Confucianism also emphasize the virtuous character and reciprocity of friendship. For Aristotle, AI chatbots or AI robots can at best be utility and pleasure friendship; they can never be genuine virtue friendship. Confucianism emphasizes authenticity in self-cultivation. Without authenticity, AI friends are bad friends; they may be harmful to self-cultivation, and one should get rid of this relationship. Indeed, empirical studies show that AI friendship is addictive. It may lead to manipulation and further social isolation. Thus, proper regulations to safeguard autonomy, emotional health and social connection, and AI ethics education are called for.