MANIPULATIVE STRATEGIES OF MODERN ARTIFICIAL INTELLIGENCE MODELS: THE PROBLEM OF UNCONTROLLABILITY AND WAYS TO ENSURE SAFE FUNCTIONING

Abstract

The contemporary development of generative artificial intelligence is characterized by the rapid expansion of its applications across all spheres of social life. Large language models and diffusion architectures are being integrated into recommendation systems, social networks, corporate communications, and political processes, acquiring an increasingly higher degree of autonomy. These systems are no longer limited to performing clearly defined tasks but demonstrate the ability for emergent properties to arise at certain levels of complexity. Such proliferation transforms artificial intelligence into an active participant in shaping public opinion, directly influencing the perception of reality by both individuals and society as a whole.