If the world's government and AI leaders ever agree to pause frontier AI training, is an international pact really feasible? How would it be implemented, coordinated and monitored?