Due to the difficulty in collecting injury cases in the actual life of intelligent service robots, it is impossible to use existing product injury information to construct injury scenarios required for risk assessment. Based on a large number of consumer and expert questionnaires, potential hazards of intelligent service robots have been preliminary identified, mainly including small component water bombs, laser radiation, and kinetic energy hazards. Using virtual reality (VR) experiments to verify and confirm the sources of harm, in order to enhance the scientific determination of product safety injury probability and severity, the weight method is used to assign different weights to the probability and severity of injury occurrence. VR experiments account for 0.4, consumer research accounts for 0.3, and expert research accounts for 0.3. Combined with matrix method, the safety risk level of the product is calculated as serious risk.