Pseudo-Consciousness in AI: Bridging the Gap Between Narrow AI and True AGI

Pseudo-consciousness bridges the gap between rigid, task-driven AI and the elusive dream of true artificial general intelligence (AGI). While modern AI excels in pattern recognition, strategic reasoning, and multimodal integration, it remains fundamentally devoid of subjective experience. Yet, emerging architectures are displaying behaviors that look intentional—adapting, self-monitoring, and making complex decisions in ways that mimic conscious cognition. If these systems can integrate information globally, reflect on their own processes, and operate with apparent goal-directed behavior, do they qualify as functionally conscious? This paper introduces pseudo-consciousness as a new conceptual category, distinct from both narrow AI and AGI. It presents a five-condition framework that defines AI capable of consciousness-like functionality without true sentience. By drawing on insights from computational theory of mind, functionalism, and neuroscientific models—such as Global Workspace Theory and Recurrent Processing Theory—we argue that intelligence and experience can be decoupled. The implications are profound. As AI systems become more autonomous and embedded in critical domains like healthcare, governance, and warfare, their ability to simulate awareness raises urgent ethical and regulatory concerns. Could a pseudo-conscious AI be trusted? Would it manipulate human perception? How do we prevent society from anthropomorphizing machines that only imitate cognition? By redefining the boundaries of intelligence and agency, this study lays the foundation for evaluating, designing, and governing AI that seems aware—without ever truly being so.