Santos, Nuno2026-06-092026-06-092026-03-1697524a9a-a3b0-4fef-bfe5-021c82585ab9http://hdl.handle.net/10400.14/58069This article introduces "Epistemological Infrastructure Capture" as a central concept to analyze how Artificial Intelligence (AI) systems increasingly mediate the infrastructures through which relevance, legitimacy, and intelligibility are socially organized. This essay argues that under these conditions, AI can synchronize perception and steer collective cognition at scale, making propaganda, radicalization, and ideological normalization structurally more effective. This article integrates Martin Heidegger's concept of Gestell (Enframing) [3] and the historical notion of Gleichschaltung (forced coordination) to illuminate how AI, particularly Large Language Models (LLMs) and recommendation algorithms, constitutes an "Algorithmic Gestell." This framework enables what this article calls an “invisible synchronization” of thought, leading to a profound crisis of epistemological sovereignty. This article differentiates between the impacts of generative and classification AI and presents data poisoning as a privileged case within this broader context. Through a philosophical and technical lens, this article explores the ontological, sociotechnical, and security implications, advocating for robust defenses and a critical re-evaluation of AI's role in shaping collective understanding.engEpistemological infrastructure captureAlgorithmic gestellData poisoningAI propagandaGleichschaltungHeideggerCognitive colonialismLLMsRecommendation algorithmsEpisemological sovereigntypreprint10.5281/ZENODO.19052972