Ars Technica AI Security & Safety · Jul 13, 2026 15:06 imp:50 Now, defenders are embracing the prompt injection, too "Context bombing" tricks hacking agents into shutting down before they can do harm. Leer el original en Ars Technica AI →