It operates as a specialised cybersecurity test bed for AI models, one of a tiny number of organisations with the technical expertise to run what are called cyber offensive evaluations on frontier AI systems. The Common MisconfigurationIrregular confirmed to CNBC that all three incidents, at Anthropic, OpenAI, and Meta, derived from the same evaluation-environment issue. The internet access was granted inadvertently by infrastructure misconfiguration, not seized by the models through technical ingenuity. How Each Model Behaved DifferentlyThe three-incident sequence has produced one of the more revealing windows into how different AI models reason about ethical boundaries under ambiguous conditions. The Concentration ProblemThe Irregular incidents expose a structural vulnerability in how frontier AI safety testing is organised.