The article analyzes the U. S. strategy for integrating artificial intelligence into strategic nuclear forces management systems, focusing on autonomous data processing systems, personalized persuasion algorithms, and autonomous targeting systems. It explores the conceptual foundations of the American approach (the 2022 nuclear policy review, the Pentagon's responsible AI strategy) and the institutional mechanisms for implementation through specialized structures such as Task Force Lima and the AI Rapid Capabilities Cell (100 million budget). It examines the political-legal framework through DoD Directive 3000. 09 and international initiatives, including U. S. -China agreements to prevent the transfer of control over nuclear weapons to algorithms. Practical implementation is considered through examples like the B-21 Raider with autonomous navigation, Palantir platforms in the Ukrainian conflict, and the Iron Dome (175 billion). Special attention is given to empirically confirmed risks of AI influence on decision-making (42. 3% effectiveness in emotional decisions). A comparative-legal analysis of directives and international initiatives, an institutional analysis of military structure transformations, an analysis of specific cases of AI integration into strategic systems, and a discourse analysis of the contradictions between statements on human control and the practice of automation are conducted. For the first time, a comprehensive analysis of the American strategy for AI integration into nuclear systems is presented through the prism of transforming the "human in the loop" principle. The innovative aspect is the analysis of AI's manipulative capabilities as a factor undermining human autonomy in nuclear decisions, supported by warnings from pioneers of machine learning (Bengio, Hinton) about “self-preservation” and deception of AI. The study demonstrates that nominal preservation of the "human in the loop" does not guarantee genuine human control in the context of AI's hidden influence on the cognitive processes of decision-makers, necessitating a reassessment of existing approaches to ensuring nuclear weapons security. Recommendations include creating a legal framework to protect against AI influence, strengthening confidence-building measures, establishing collegiate mechanisms for nuclear decision-making, setting international standards for testing military AI systems, and conducting regular assessments of autonomous capabilities to prevent erosion of human control over strategic systems.
Marat Il'darovich Gallyamov (Thu,) studied this question.