Delaying Internal Deployments Until After X Days of Safety Testing

A proposal on LessWrong suggests postponing internal AI system deployments until they have undergone a specified number of days of safety testing to enhance reliability and risk management.