Responsibility is not required for authorship

The Committee on Publication Ethics (COPE) maintains that AIs (artificial intelligences) cannot be authors of academic papers, because they are unable to take responsibility for them. COPE appears to have the answerability sense of responsibility in mind. It is true that AIs cannot be answerable for papers, but responsibility in this sense is not required for authorship in the sciences. I suggest that ethics will be forced to follow suit in dropping responsibility as a criterion for authorship or rethinking its role. I put forward three options for authorship: dropping responsibility as a criterion for authorship, retaining it and excluding AIs, but at the cost of substantial revision of our practices, or requiring only local responsibility for an intellectual contribution.