Attack technique against AI systems and Large Language Models that exploits malicious inputs to manipulate model behavior, bypass restrictions, extract sensitive data, or perform unauthorized actions through prompt crafting and context poisoning.