The advancement of artificial intelligence has led to exploration across various fields, including transportation. Traffic risk assessment has emerged as a prominent research focus within this domain, driven by the need to enhance road safety and improve driver assistance systems. Researchers are increasingly leveraging AI and deep learning techniques to develop sophisticated models that can accurately predict and classify traffic risks, providing valuable insights for intelligent transportation systems. This growing interest underscores the critical role that AI can play in mitigating traffic-related hazards and enhancing overall road safety. The objective of this paper is to classify traffic risks based on visual scenarios recorded by dashcam videos. We propose in this paper an efficient CNN model for traffic risk classification into critical risk, high risk, low risk, and moderate risk using a traffic risk assessment dataset. Our model results are very efficient, surpassing all models previously proposed in the literature for this objective. Our achieved model has an accuracy of 99%, a loss function value of 0.03, and a ROC AUC score of 99.25%.