An apparatus for artificially and intelligently identifying money over and short in an RFID cashbox is disclosed. The apparatus comprises certificates of deposit and withdrawal, a cabinet box, a counter group, a foot box, a camera shooting system, a damping keyboard, a character identification system and a financial server. The cabinet box is used for placing the counter group. The counter group is formed by a plurality of counters and is used for placing scattered money and bundles of money. Each counter comprises an electromagnetic force balance weighing sensor located in a lower portion anda counter body located in an upper portion. The electromagnetic force balance weighing sensor comprises a bearing bench located in an upper portion and a pedestal located in a lower portion. A pedestal bottom is provided with a circuit interface. An artificial intelligence fuzzy identification server uses a neural network image identification algorithm, acquires all or parts of images of a specific face value of money from the camera shooting system and assesses a condition of the specific face value of money. A fuzzy k-nearest neighbor algorithm is used. The face value of each piece of moneyin the same counter is assumed to be the same. Because of the different conditions of the money, weights are different. A fuzzy rule base is set for each counter. According to the weights and added weights, the number of the money is determined.