Artificial Intelligence (AI) has entered all walks of life and should be developed on the basis of security. Large models of Artificial Intelligence can constantly learn, have thinking ability, reasoning ability and human-like "consciousness", can express views and attitudes, and can reflect the "cognition" of the world, which may produce errors, biases and risks in its decision-making and information processing, and then affect human cognition and thought. If Artificial Intelligence does not show a correct cognition in the value orientation, if it is improperly and unreasonably used, it may have an adverse impact on various aspects of society, economy, culture and people's lives. Furthermore, the cognitive errors and deviations of Artificial Intelligence may cause problems such as tearing, manipulation and deconstruction of correct social cognition, resulting in serious social risks and even social chaos and turmoil, so reasonable regulations should be carried out on Artificial Intelligence. The background of this study is the rapid integration of Artificial Intelligence into decision-making, information processing, and daily interactions necessitates a critical examination of its impact on human cognition. This study aims to systematically identify, categorize, and analyze the primary cognitive risks associated with human-AI interaction. The research methods employed in this article include literature research method, case study method, and interdisciplinary research method, etc. This study provides empirical evidence confirming significant cognitive risks inherent in human-AI interaction. Proactively addressing these cognitive risks is crucial for harnessing AI's benefits while safeguarding human cognitive autonomy and decision-making integrity.

Artificial Intelligence, Consciousness, Risk