I believe AI labs should pause training when AI Safety Institutes detect signs of potential loss of control. Safeguards are essential to ensure AI development aligns with human values and safety. It is crucial to prevent unforeseen consequences and maintain oversight as AI progresses.