Social power is everywhere. It is not only the fundamental force in social relations but also one of the most fundamental concepts in social science. Attribution refers to how people provide a causal explanation by speculating on others’ behaviors and mental status. It is crucial to understanding and predicting the environment. As an important research topic of social cognition, the relationship between attribution and social power has attracted increasing attention in academia. However, as a fundamental force of social relations, how social power affects attribution errors remains to be studied. According to the dual-process model, there are two different processes. One is an automatic process that is fast, instinctive, and intuitive. People make judgments unconsciously, based on simple rules. The other process is the controlled process. This cognitive process is a relatively slow, sequential process that requires conscious effort. The theory of the two-step attribution process holds that the automatic process generates internal attribution, and only further, controlled processes can lead to external attribution. Enhancing social power results in automatic social cognition, and vice versa, leading to controlled social cognition. This assertion of the approach/inhibition theory of social power has been validated in several studies. So, social power may increase the tendency toward automatic processes, thus raising an underlying attribution error. We verified this theory using four studies. The first three studies use the sustainable traits of power. Study One used situational tasks to measure the fundamental attribution error; Study Two and Study Three adopted questionnaires to evaluate the tendency of internal and external attributions. Additionally, Study Two and Study Three set socioeconomic status as a control variable due to the association between socioeconomic status and social power. Study Four initiated high and low social power roles through priming experiments to examine the causal relationship between social power and internal/external attributions. All four studies have reached a consistent conclusion: social power increases the fundamental attribution error. There was a significant positive correlation between social power and attribution to high fuzzy events in Study One, but not in medium fuzzy situation or low fuzzy situation. In study Two, after variables such as gender and socioeconomic status were controlled for, social power remained a significant predictor of the wealth gap attribution: the higher the social power was, the more likely an internal attribution would be made to explain the wealth gap. This finding agreed with our research hypothesis. Study Three shows again that social power could positively predict the tendency to commit the fundamental attribution error.The effects of Study One and Study Four were different in high fuzzy situations. In Study One, the impact of the difference in the fundamental attribution error between the high and low social power groups was small. This effect, however, was moderate in Study Four. This discrepancy may be because Study One used a questionnaire with many confounding variables that were not well controlled. In contrast, Study Four was experimental research that was conducted with random grouping, and therefore, the confounding variables were better managed in the experiment.