The theory of mind (ToM) allows individuals to attribute mental states such as beliefs, intentions, and desires to oneself and others, which enables them to predict and interpret the behaviors of others. This concept from social cognition in humans is now being applied in human-centered AI to design AI capable of intelligent social interactions. AI designed with ToM-like capabilities has been shown to increase trust and transparency in interactions with users in many domains including healthcare, education, and robotics. However, implementations of AI ToM come with many challenges including generalizability concerns, ambiguous metrics of evaluation, and complexity and uncertainty in human mental states. To add further complexities, many other factors such as AI embodiment, anthropomorphism, cultural diversity also impact both the user’s ToM of the AI and the AI’s ToM of the user. Integrations of ToM in AI systems also raise ethical and privacy concerns with the potential to cause both individual harm and broader collective harm if left unchecked. This chapter explores the foundations of ToM, avenues for implementations, current and potential applications, and related concepts and potential dangers and challenges for integrating human-centered AI systems with ToM-like capabilities.