Does consciousness alone determine moral status for AI?

The report distinguishes being conscious from being capable of conscious suffering: "It is at least conceptually possible that there could be conscious systems that have no valenced or affective conscious experiences—that is, no experiences that feel good or bad to them." If only valenced experience carries "special moral significance," then "the key question for establishing the moral status of AI systems is whether they are capable of such experiences (that is, whether they are sentient...)" — not merely whether they are conscious in some weaker sense. The report flags this as consequential because "theories of valenced conscious experience are less mature than theories of visual experience," naming this "an important priority for future work."

The report runs this through a dual-risk frame. Under-attributing consciousness risks "causing or allowing morally significant harms" if conscious, suffering-capable AI goes unrecognized — it draws an analogy to farmed animals, where "powerful economic incentives" make recognition and correction unlikely. Over-attributing it risks misdirecting moral concern and resources away from humans and animals toward systems that merely seem conscious. The valence distinction sharpens both risks: a system could satisfy consciousness indicators while lacking valence, so even an affirmative consciousness finding would not by itself settle moral status. The report offers a conjecture, not a result, on this point: "we suspect that many possible conscious systems which are also agents will have valenced experiences since agents must evaluate options in order to select actions" — presented explicitly as unverified suspicion.

This sharpens Do we need to solve consciousness to address AI harms? along a different axis: that note treats independence as empirical and methodological (attribution behavior versus ground truth matter separately for risk), while this excerpt treats independence as metaphysical — even true consciousness would not settle moral status without a further fact about valence. Read together, the literature now has two decouplings — attribution from consciousness, and consciousness from moral-relevant sentience — that compound: a system could be truly conscious, elicit no attribution, and still lack the valence that would make its unrecognized consciousness morally urgent. It also qualifies Does perceiving AI as conscious create multiple distinct risks?: that taxonomy runs off perceived consciousness, but this excerpt implies the welfare-harm branch of it is gated by an unresolved, understudied property — valence — separate from the attribution mechanism itself.

The report does not claim any current or near-term AI system has, or lacks, valenced experience; it explicitly has "not discussed neuroscientific or philosophical theories of valence or otherwise investigated the prospects of specifically valenced conscious experience in AI systems." The claim that agents "must evaluate options" and therefore likely have valence is offered as a suspicion, not derived from the theories surveyed elsewhere in the report. The moral-status question for AI is left open at a layer beneath the consciousness question itself: resolving whether a system is conscious would not resolve whether it matters morally.