Have we finally solved social engineering? Plus: World Cup fraud, AI IDs and an IBM/OpenAI collab

AI-native operating systems with integrated Large Language Models (LLMs) could potentially solve social engineering by providing context-aware, automated defenses that remove humans from routine trust decisions, addressing three key human limitations: lack of context for security decisions, cognitive overload from attackers, and poor phishing detection. However, this solution faces challenges including LLM vulnerability to prompt injection attacks, scalability issues with millions of AI agents requiring identity management, and the need for guardrails to prevent AI agents from becoming attack vectors themselves. The shift from human-based to AI-based social engineering attacks means defenders must now protect both humans and AI systems.