Autonomous driving is still plagued by threats from various attacks, which can be divided into physical attacks, network attacks, and learning-based adversarial attacks. Inevitably, the safety & security of autonomous driving systems based on deep learning are seriously challenged by these attacks, so countermeasures should be studied to mitigate potential risks. The current open challenges in the safety of autonomous driving include the following points. Section 11.1 introduces some basic safety concepts, Section 11.2 analyzes NHTSA’s 12 safety elements, Section 11.3 introduces the functional safety standard ISO 26262, Section 11.4 discusses the expected functional safety standard SOTIF, Section 11.5 analyzes Mobileye’s safety model RSS, Section 11.6 introduces network security, and Section 11.7 elaborates on the safety hazards of autonomous driving and solutions.