The rapid development of large language models has made the question of whether AI can be conscious no longer merely a philosophical issue, but also a practical concern in technology ethics and governance. Alexander Lerchner's concept of the "abstraction fallacy" attempts to fundamentally challenge the possibility that current large language models could lead to genuine consciousness. At its core, AI computation is a symbolic processing of the world: it can simulate human language, emotions, and behavior, but this does not mean that it truly generates subjective experience. This article argues that AI consciousness will not naturally emerge simply from larger models, greater computational power, or more complex algorithms. AI may increasingly resemble human beings; it may say "I am in pain", "I am afraid", or "I hope", and it may display complex communication and planning abilities. Yet these remain forms of behavioral imitation and functional simulation, and they do not demonstrate that AI truly has feelings. Current and foreseeable artificial intelligence systems should therefore be regarded as powerful non−conscious tools rather than subjects with subjective experience. The focus of future AI governance should not be a premature discussion of machine rights, but the prevention of excessive anthropomorphism, responsibility displacement, and the underestimation of real risks posed by highly capable AI tools.