Abstract: AI-assisted driving systems introduce the risks of technical defects and human-machine interaction errors, both of which can lead to traffic accidents. When a technical defect in an AI-assisted system becomes a reality and causes an accident, AI service providers and car manufacturers may face charges of negligent injury or negligent manslaughter, but not for traffic offenses. In determining negligence, it should be assessed from an ex-post perspective regarding the risks and causes of technical defects. However, from an ex-ante perspective, the foreseeability of technical defects should be judged based on existing technology and permissible risk principles. When human-machine interaction errors occur and result in accidents, drivers may bear criminal liability for traffic offenses. In assessing negligence, the principle of trust should be reasonably applied to define the scope of duty of care, and the basic principles of negligence for omissions should be used to determine whether the driver has violated the duty to take control. In cases involving joint negligence by AI service providers, car manufacturers, and drivers, the scope of criminal liability for negligence of AI service providers and car manufacturers should be limited by referring to the elements of traffic offense liability.