The question of whether advanced AI systems might deserve moral consideration is deeply uncomfortable, but the cost of getting it wrong in either direction is enormous. Is it premature to discuss AI welfare, or is now exactly the right time — before systems exist that might qualify?