Deep uncertainty surrounds whether artificial systems can ever be conscious, leaving us unsure how to treat AI that might be sentient. On one side, assuming insentience risks causing grave harms to entities that deserve moral standing; on the other, assuming sentience may waste resources on machines that lack consciousness. Because the core questions about AI consciousness are intractable in the near term, this dilemma is hard to escape. I therefore propose shifting the focus from the elusive notion of consciousness to the more tractable concept of AI valence. Specifically, we can examine whether an AI exhibits states that would correspond to valenced experiences if it were conscious. Demonstrating the presence of such states provides a sufficient ethical grounding for responsibly developing AI that could be conscious.\n\nBlogger's Review: By moving the ethical discussion from the abstract and hard‑to‑verify realm of consciousness to the observable domain of valence, this approach offers a pragmatic and actionable framework for AI research and policy.