This paper presents a systematic review of explainable AI (XAI) methods for safe and trustworthy autonomous driving (AD). The review identifies five key contributions of XAI for AD: interpretable design, interpretable surrogate models, interpretable monitoring, auxiliary explanations, and interpretable validation. The authors propose a conceptual framework called SafeX to integrate these methods, enabling explanation delivery to users while ensuring AI model safety. The review highlights the importance of XAI in meeting the requirements for AI in AD, including data, model, and agency. The authors also discuss the sources of explanations in AI systems and introduce a taxonomy of XAI concepts. The review finds that XAI is fundamental to meeting the requirements for safe and trustworthy AI in AD. The authors also discuss the challenges of integrating XAI with AD and propose a conceptual framework called SafeX to address these challenges. The review concludes that XAI is an essential tool in meeting the requirements of safe and trustworthy AI for AD.This paper presents a systematic review of explainable AI (XAI) methods for safe and trustworthy autonomous driving (AD). The review identifies five key contributions of XAI for AD: interpretable design, interpretable surrogate models, interpretable monitoring, auxiliary explanations, and interpretable validation. The authors propose a conceptual framework called SafeX to integrate these methods, enabling explanation delivery to users while ensuring AI model safety. The review highlights the importance of XAI in meeting the requirements for AI in AD, including data, model, and agency. The authors also discuss the sources of explanations in AI systems and introduce a taxonomy of XAI concepts. The review finds that XAI is fundamental to meeting the requirements for safe and trustworthy AI in AD. The authors also discuss the challenges of integrating XAI with AD and propose a conceptual framework called SafeX to address these challenges. The review concludes that XAI is an essential tool in meeting the requirements of safe and trustworthy AI for AD.