The objective of this study is to verify the robustness of Decision Transformer, an offline reinforcement learning model using Transformer architecture, in robot control. Decision Transformer has been reported to perform better than traditional reinforcement learning methods. In this study, four different robots—Half Cheetah, Hopper, Walker2D, and Ant—were used to evaluate failure cases not included in the collected data. Robustness was evaluated using three datasets (Medium, Medium-Expert, and Medium-Replay) to simulate situations where the actuators malfunctioned. Experimental results demonstrated a trend of lower rewards across all robots and datasets, indicating a lack of robustness in Decision Transformer. These results highlight the need to enhance the robustness of Decision Transformer for robust robot control.