The accuracy of facial recognition technology is not that it performs poorly for certain races, but that there may be different degrees of deviation in different groups, but this is not based on the inherent characteristics of the race itself, but caused by a variety of complex factors. Studies have shown that facial recognition technology has differences in different genders, ages, skin tones, facial expressions, lighting conditions, and facial hair. For example, in low-light environments, the facial recognition accuracy rate of dark-skinned people may be relatively low. This is due to the algorithm's imperfect handling of skin color differences, rather than discriminatory accuracy defects for a certain race. In the design and training process of facial recognition technology, there may be uneven data. The training data often does not fully represent the facial features and changes of all groups of people, especially the historical data has less data for certain groups, which leads to poor algorithms in identifying these groups. This kind of technical deviation has brought about a series of problems, which may lead to certain groups facing unfair treatment in the fields of security monitoring, law enforcement, finance, etc. Some countries and regions have become aware of this problem and have taken measures to improve algorithms and increase data diversity to improve the fairness and accuracy of facial recognition technology and reduce potential bias effects.