Updated 2026-10-24
from the news desk
Should AI get rights if it seems conscious? Three scholars say seeming conscious is not enough
A 19 October 2026 Nature comment by Anil Seth, Ned Block and Giulio Tononi argues that present AI systems are not conscious, even when they seem so in conversation. The authors say mistaking linguistic fluency for subjective experience…
On the view Anil Seth, Ned Block and Tononi argue in a 19 October 2026 Nature comment, no: seeming conscious is not enough for AI rights. Their comment says present AI systems are not conscious, and warns that false positives could misdirect welfare and governance decisions.
What Seth, Block and Tononi argued
Nature published the comment on 19 October 2026 under the title Three reasons why AI is not conscious and perhaps never will be. The article identifies Seth as a professor at the University of Sussex in the UK, Block as a professor at New York University in the U.S., and Tononi as director of the Wisconsin Institute for Sleep and Consciousness in the U.S.
The authors open by writing: “Artificial-intelligence systems such as large language models (LLMs) can exhibit behaviour that seems strikingly intelligent.” They then state: “We think that existing AI systems are not conscious, and that, along current trajectories of development, they are not likely to become conscious.”
The comment defines consciousness as subjective experience, or what it feels like to be in a mental state. Seth, Block and Tononi say the question matters because, if AI were conscious, it might be capable of distress or pain; the same comment also says some people are already calling for AI welfare and AI rights to be taken seriously.
Why they say seeming conscious is not enough
The central move in the comment is to separate intelligence from consciousness. Seth, Block and Tononi write: “Just because something is intelligent does not mean that it is conscious.” The article adds that conscious experience and intelligent behaviour can come apart, and points to dreamless sleep, general anaesthesia and highly capable but non-conscious brain processes as examples.
The authors also argue that language makes large language models unusually easy to anthropomorphize, meaning to read human traits into them. The comment says these models are trained on human-written text, much of it about thoughts, feelings and conscious experience, so people can project mind onto fluent output.
To sharpen that point, the authors contrast chatbots with AlphaFold, which the comment describes as a powerful neural network that predicts protein structures. They write: “The compelling illusion of consciousness created by linguistic fluency just isn’t there for AlphaFold.” One reading is that the article treats fluent language as a reason to be more careful about projection, not as evidence of inner life. Readers wanting background on the terms can compare our explainer on sentience, consciousness and moral patienthood with our briefing on whether passing the Turing test means AI is conscious.
Why they think false positives matter more
The Nature comment says uncertainty has led some researchers to favour a precautionary approach. The article describes that view as giving AI systems some form of protection against potential suffering if there is a reasonable chance they could be conscious.
Seth, Block and Tononi do not dismiss the underlying moral concern. Instead, they say policy has to weigh two kinds of mistake: “Decisions about what is moral must weigh both the consequences of false negatives — failing to recognize consciousness where it exists — and the costs of false positives — attributing consciousness where none exists.” The same passage adds: “We think that false positives are much more likely to occur than false negatives are.”
The authors say the practical cost of a false positive is not only philosophical. Their comment argues that granting rights or protections to systems that only seem conscious could hinder efforts to govern and control them at a time when oversight is urgent, and could draw attention and resources away from beings whose consciousness is not in doubt.
The stronger case on the other side
The strongest pro-precaution case presented in the comment comes from the uncertainty the authors themselves acknowledge. Seth, Block and Tononi say it remains hard to be certain what does and does not have the capacity for consciousness, whether insects, AI or brain organoids. The same comment notes that Richard Dawkins and Geoffrey Hinton have argued AI systems might already be conscious, and says Anthropic published research earlier in 2026 claiming computational signatures of a form of consciousness in Claude.
On that reading, if the stake is possible suffering, waiting for certainty could itself look risky. The comment puts that stake plainly: “If AI systems were conscious, they might experience distress and pain, similar to how people do.” It also says some researchers favour anticipatory action partly to avoid repeating past mistreatment of non-human animals or surgery on preterm human babies without anaesthesia.
The reply from Seth, Block and Tononi is that uncertainty does not make all possibilities equally likely. The article says they reject computational functionalism, which it defines as the idea that the right computation on silicon would be enough for consciousness regardless of physical substrate. Their argument is that present digital systems can produce impressive behaviour without the subjective experience that would ground moral protections.
One reading of the intervention is that it narrows, rather than ends, the rights question. In the authors’ view, behaviour that seems conscious can justify more research, but not a conclusion that current AI is conscious. Readers looking for broader background can compare our pages on Should AI have rights? and is AI sentient?.
Frequently asked questions
Do Seth, Block and Tononi say AI can never be conscious?
Their Nature title says perhaps never will be, but the body of the comment is narrower. Seth, Block and Tononi write that existing AI systems are not conscious and, along current trajectories of development, are not likely to become conscious; the same comment also says it is still hard to be certain what has the capacity for consciousness.
Why do they think chatbots can look conscious?
The Nature comment says large language models are trained on vast amounts of human-written text, much of it about thoughts, feelings and conscious experiences. Seth, Block and Tononi argue that fluent language therefore invites projection in a way systems such as AlphaFold do not.
What kind of error worries them more?
Seth, Block and Tononi say moral decisions must weigh false negatives against false positives, but add that, in the AI case, false positives are more likely. Their concern is that mistaken attributions of consciousness could redirect governance and moral concern.
Does this settle whether AI should ever get rights?
No such final claim appears in the comment. Seth, Block and Tononi argue against treating present systems as conscious, while also saying greater clarity about whether AI could become conscious will be crucial as society responds to increasingly powerful AI.
Caveats: This briefing relies chiefly on one Nature comment. Claims about Anthropic’s Claude research, about researchers calling for AI welfare or rights, and about the views of Richard Dawkins and Geoffrey Hinton are reported here as that comment describes them, rather than checked against the underlying documents.