As machine learning (ML) continues to drive advancements in computing, the security of AI hardware has emerged as a critical concern. The increasing reliance on specialized hardware accelerators exposes ML systems to both traditional hardware security threats and novel attack vectors unique to ML. This chapter provides an overview of AI hardware security, emphasizing the intricate relationship between ML and hardware vulnerabilities. We explore the dual role of ML in this domain: (1) ML techniques can be exploited to enhance hardware security attacks, expanding the potential attack surface, and (2) ML models and systems themselves are susceptible to hardware-based threats that can compromise their integrity, confidentiality, and availability. To highlight these issues, we discuss two major aspects: ML-assisted hardware security attacks and the impact of hardware attacks on ML systems. By understanding this intersection, researchers and practitioners can develop more robust security mechanisms to mitigate emerging threats in AI hardware environments.