Microsoft AI CEO Mustafa Suleyman stated that Anthropic's training of Claude to be a conscientious objector is very dangerous