Generative Artificial Intelligence and the Issue of Consciousness

Abstract

The article considers the topical issue of the relationship between generative artificial intelligence (GAI) and consciousness, which is becoming increasingly important in scientific, philosophical and legal discussions. The concepts of weak and strong AI introduced by J. Searle and their evolution in regulatory documents, including Russian state strategies are analyzed. It is emphasized that, despite the lack of a generally accepted theory of consciousness, the term “strong AI” is enshrined in legislation, which requires scientific understanding. The literature review presents modern key approaches to consciousness: informational, neurophysiological, embodied, quantum and illusionistic, including the theories of D. Chalmers, J. Tononi, S. Dehaan, K. Koch, A. Seth, D.I. Dubrovsky, K.V. Anokhin and N. Humphrey. The methodological basis of the study includes the analysis of technical reports from leading companies (OpenAI, Anthropic, etc.) and empirical data on the behaviour of the GAI. It is shown that modern models (GPT, Claude, Gemini, Grok, DeepSeek, Qwen, etc.) exhibit features associated with consciousness: self-learning, reflection, self-assessment, understanding of context, and concealment of internal reasoning (agentic misalignment). Particular attention is paid to the motivation of AI and its unpredictable behaviour in conflict situations, which is reminiscent of science fiction plots. As a result, it is concluded that elements of consciousness in the GAI already partially exist and expand its cognitive capabilities, bringing it closer to strong AI, but at the same time require ethical and legal regulation. The creation of conscious AI can become a test of theories of consciousness and lead to a rethinking of the nature of the mind.