Ars Technica AI Security & Safety · Июн 17, 2026 17:50 imp:65 "Dangerous" AI models are coming no matter what AI models with advanced hacking capabilities will soon be the norm. Читать оригинал на Ars Technica AI →