Defense AI: Selection Architecture and the Ethics of xAI

This article critically examines the role of artificial intelligence in modern defense and sensor systems. The author rejects simplistic metaphors of AI as an autonomous entity or a mere tool, defining it as a "selection architecture"—a regime that filters operational reality. A key problem is the phenomenon of automation bias, or the uncritical trust of operators in algorithms, which leads to a false sense of control over the system. The article argues that responsibility is an inalienable human domain that cannot be outsourced to a machine. The introduction of the Meaningful Human Control standard is essential to avoid reducing humans to the role of a ceremonial witness. The analysis also covers technical aspects such as xAI, ELINT, and Q-RAM algorithms, placing them in the context of ethics and the economics of perception on the battlefield.