Змагальний штучний інтелект в інформаційних і кібернетичних війнах

This article is dedicated to the role of adversarial artificial intelligence (Adversarial AI) in modern hybrid conflicts and their integral informational, cybernetic, and legal components. Adversarial AI is examined as a manifestation of conflict between artificial intelligences (AI) within the context of the authors' proposed foundations of AI conflictology. The study explores cyber threat models that describe the dynamics of offensive and defensive strategies in cyberspace, including simulations of various attack types and the development of mechanisms for their neutralization. Attention is given to information warfare, analyzing the impact of manipulative content on audiences and developing methods for its detection, analysis, and blocking. The use of neurolinguistic programming in adversarial AI is also considered. Additionally, models for detecting and neutralizing backdoors in large language models are examined in the context of adversarial AI. The proposed model allows for analyzing the effectiveness of creating and implementing backdoors, as well as improving methods for their search, detection, and neutralization. The article also discusses adversarial artificial intelligence in the context of its application in informational and cyber conflicts from a legal perspective, analyzing the role of international regulation in controlling the development and use of such technologies. Special attention is paid to issues of responsibility for the misuse of these technologies, defining the boundaries of lawful use, and mechanisms for countering illegal cyber operations.