Against the dual backdrop of the deepening integration of artificial intelligence into society and the accelerated development of an age-friendly society, AI literacy has acquired new connotations. Among older adults, disparities in AI literacy have extended beyond behavioral and operational deficiencies in skills to the level of value-based reasoning, manifested in dependence and addiction, fear and rejection, and cognitive simplification and rigidity. From the perspective of bounded technology governance, the inadequate and inappropriate integration of people and technology is the deep-seated cause of these disparities. This study proposes that science popularization on AI literacy for older adults should be grounded in four fundamental principles: a people-centered orientation; practicality and moderation; systematicity and scientific rigor; and inclusiveness and universal benefit. It should focus on cognitive reconstruction and value formation while appropriately adjusting the subjects, content, methods, and evaluation system of science popularization. The aim is to help older adults take the initiative in controlling technology, become co-creators of meaning and guardians of values in an AI-assisted society, and escape both the traps of AI and the marginalization experienced by "AI refugees."