While AI safety is crucial, requiring labs to pause training based on external evaluations could stifle innovation and progress. We need a balanced approach that encourages responsible development instead of strict regulation.