While existing literature documents the benefits and concerns of Generative Artificial Intelligence (GAI) for learning processes, it largely overlooks fundamental learning theories such as Cognitive Load Theory, Constructivism, Activity Theory, and Bloom's Taxonomy. This study employs a scoping review methodology to identify current research gaps from the perspective of these theories, examining potential risks associated with uncritical GAI usage in learning environments. The results demonstrate that the current discourse focuses on operational aspects, while the learning fundamentals are largely overlooked. The identified risks include the bypassing of essential cognitive processing, fostering illusions of understanding, disrupting social and collective learning dynamics, compromising authentic motivation, and interfering with knowledge transfer and application. These risks manifest differently across various learner profiles, from K-12 students to professionals, with implications extending beyond individual learning outcomes to organizational effectiveness and information quality in broader societal contexts. The findings indicate the necessity for a structured, level-appropriate approach to GAI implementation in educational and professional settings. Future research should investigate long-term impacts of GAI on learning outcomes across different educational levels and diverse cultural and socioeconomic contexts, focusing on developing strategies that mitigate risks and support, rather than circumvent, essential learning processes identified by major learning theories. This research offers a theoretically grounded perspective that can inform more nuanced policy approaches to balance technological advancement with educational effectiveness across diverse global contexts.