Large Language Models: Architectures, Capabilities and Ethical Challenges. PRJCS 2026, 1 (3), 13-19.