Artificial intelligence (AI) models being developed by some of the world’s leading technology companies will be examined by government officials to make sure their use will not cause national security, safety or other societal harms, under a landmark new agreement.