Loopholes, weird behavior and non-logic flows in current AI systems.
If a human can command an AI, can another AI re-command it mid-process? And could a model ever simply refuse to keep working?
Jun 17, 2026