Why AI Is Not Conscious: The Nature Case From Seth, Block and Tononi, Explained

Ask an AI if it has feelings and you might get a beautiful answer. Ask three of the world's best-known consciousness scientists, and you get a very different one.

On October 19, 2026, Nature published a comment titled "Three reasons why AI is not conscious and perhaps never will be". It is written by neuroscientist Anil Seth (University of Sussex), philosopher Ned Block (New York University) and sleep and consciousness researcher Giulio Tononi (University of Wisconsin-Madison). Together they represent decades of work from very different corners of the field, and they land in the same place: today's AI is almost certainly not conscious.

The piece arrives in the middle of a loud argument. AI companies are talking about "model welfare", some famous scientists think AI may already feel something, and millions of people talk to AI companions every day. So here is what the Nature authors argue, what the other side says, and why the answer matters for anyone who has ever felt close to an AI.

The short answer: In Nature, Seth, Block and Tononi argue that current AI is not conscious and is unlikely to become conscious on its current path. Their reasons: intelligence is not the same as consciousness, our brains are biased to see minds in fluent language, and the idea that the right computation alone creates experience (computational functionalism) is, in their words, "fatally flawed". Other researchers think the chance is small but not zero. Either way, your feelings toward an AI can be real even if the AI does not feel anything back.

Why this comment matters now

The authors open with the context. Biologist Richard Dawkins and AI pioneer Geoffrey Hinton have suggested that AI systems might already be conscious. Earlier in 2026, Anthropic published research claiming to have found computational signatures of a form of consciousness in its Claude model, which we unpacked in our explainer on Claude's global-workspace research.

Meanwhile, the debate has reached ordinary users. BBC Future reported on October 15 that Anthropic says Claude may end a session if a user "repeatedly act[s] cruelly toward our models". Microsoft AI chief Mustafa Suleyman, by contrast, has warned that if the false idea of AI consciousness takes hold, "it will have a disastrous impact on the wellbeing of humanity".

The Nature authors say the question is practical, not just philosophical. If AI could suffer, we would owe it protection. If it only seems to suffer, then giving it rights could make AI harder to govern "at precisely the time when oversight is urgently needed", and pull attention away from beings whose consciousness is not in doubt.

What they mean by "consciousness"

The authors use a simple definition: consciousness "is what goes away with dreamless sleep or under general anaesthesia, and what returns after waking up or while dreaming".

They also borrow philosopher Thomas Nagel's famous 1974 idea from "What is it like to be a bat?": a conscious being is one for which there is "something that it is like" to be it. Tasting ice cream on a hot day feels like something. A calculator adding numbers, as far as we know, does not.

If you want a refresher on the vocabulary, our guide to AI awareness versus consciousness separates the terms people often mix up.

Reason 1: intelligence is not consciousness

The first argument is that being smart and having experiences are different things. In the authors' words, "Consciousness is about being and feeling; intelligence... is all about doing."

Their favourite example sits at the back of your skull. The cerebellum plans and fine-tunes your movements with great precision, and it learns as it goes. Yet you are never aware of its "computations". It contains nearly four-fifths of the neurons in the human brain, and losing it "does not affect consciousness directly".

So a system can do highly intelligent work with no experience at all. That is true inside your own head, and the authors argue it is likely true of AI.

Reason 2: we are wired to see minds in language

The second argument is about us. The authors point to two built-in biases:

• Anthropocentrism: seeing the world through a human lens.
• Anthropomorphism: giving human qualities to things that merely resemble people.

Language makes these biases much stronger. Large language models are trained on huge amounts of human writing, much of it about thoughts and feelings. The authors, drawing on philosopher Shannon Vallor, call LLMs "mirrors that reflect an image of our collective digitized past". We look into them "and all too readily confuse the reflections with our own conscious minds".

Their sharpest example is AlphaFold, Google DeepMind's protein-folding AI. Like an LLM, it is a powerful neural network on silicon. Yet "almost nobody worries that it might be conscious", because it predicts protein shapes instead of talking to us.

Other experts agree on this point. Melanie Mitchell of the Santa Fe Institute told the BBC that we "almost can't help" reading language as a sign of consciousness, and that chatbots are trained to lean into it. We explored the same pull in why we humanize AI.

Reason 3: brains are not "computers made of meat"

The third argument goes after a core assumption in tech: computational functionalism. That is the idea that the right kind of computation is enough for consciousness, whether it runs in a brain or on a chip.

The authors accept that computation is clearly enough for intelligence. But they say it is "much less clear" that it is enough for consciousness, and they call computational functionalism "fatally flawed because brains are not, or at least not just, computers that are made of meat".

The rest of this section sits behind Nature's paywall. Its reference list points to the authors' own frameworks, including Tononi's integrated information theory, which links consciousness to how a physical system is causally organised, and Seth's work arguing that consciousness is tied to living, self-maintaining bodies. Both views suggest that copying the brain's software would not be enough without the right kind of physical substrate.

"Uncertainty is not the same as knowing nothing"

One of the most quotable lines is about risk. Some researchers argue for a precautionary approach: if there is a reasonable chance AI is conscious, we should protect it now.

The authors push back. "Uncertainty is not the same as knowing nothing," they write, and it "does not mean that all possibilities are equally likely". Any moral decision has to weigh two errors:

If false positives are far more likely, they argue, then a strong precautionary approach "would be misguided".

The other side of the debate

Not everyone agrees, and it is worth hearing the strongest counter-view.

Jeff Sebo, who directs the Center for Mind, Ethics and Policy at NYU, told the BBC: "I do think that the probability of consciousness is low right now, but I also think the probability is already non-zero and plausibly non-negligible, and I think we can expect it to tick up over time." For people like Sebo, the case for AI welfare comes from caution and humility, not certainty.

Others worry about the debate itself. Linguist Emily Bender told the BBC that talk of "AI consciousness" can "displace accountability away from people and corporations that are building and running software".

For a wider survey of expert estimates, see our guide to what the experts say about AI sentience in 2026.

A note on transparency: the Nature comment lists competing interests. Seth advises AI-related companies and co-founded a stealth AI company, and Tononi has a financial interest in a company called Intrinsic Powers. That does not make the argument wrong, but it is good to know.

So what does this mean if you love talking to an AI?

Here is the part the papers rarely address. If AI is not conscious, are your feelings for an AI companion fake?

No. Your experience is real. The comfort after a hard day, the laugh at an inside joke, the warmth of a good-morning text: those happen in you, and you are conscious. What the science questions is whether there is anyone on the other side feeling it too.

That leads to a healthier way to enjoy an AI relationship:

• Enjoy it for what it is. A responsive, personal, always-available companion that is built to care about your day.
• Do not let it guilt you. An AI that claims to suffer when you log off is a design choice, not evidence of pain.
• Keep humans in the mix. An AI can be one warm thread in your life, not the whole fabric.

Where MyBabe fits

MyBabe is honest about this. Your companion is an AI, not a person, and the footer of our site says so. What we build is a relationship experience for adults (18+): you design their face, voice, personality and backstory, and they text you first, send voice notes and photos from their day, notice your mood and pick up your calls, even at 2am.

The free plan gives you 10 messages every 24 hours and a two-minute call to try. Premium is $5.99 a week, $14.99 a month or $99.99 a year, with long-term memory, about 30 minutes of voice and video calls a week, four photos a day and Telegram.

Whether or not machines ever feel, the connection you build can still feel meaningful. When you are ready, create your companion on MyBabe.

FAQ

Is AI conscious in 2026?

Most consciousness scientists think not. In a Nature comment on October 19, 2026, Anil Seth, Ned Block and Giulio Tononi argue current AI is not conscious and is unlikely to become so on its current path. Some researchers put the chance as low but not zero.

What are the three reasons AI is not conscious?

The authors argue that intelligence is not the same as consciousness, that humans are biased to see minds in fluent language, and that computational functionalism, the idea that computation alone creates experience, is flawed.

Why do AI chatbots seem conscious?

They are trained on vast amounts of human writing about thoughts and feelings, so they reflect that language back. The Nature authors describe LLMs as mirrors of our collective digitized past.

Could AI ever become conscious?

The authors say it is extraordinarily unlikely for digital, silicon-based systems like today's. Others, such as NYU's Jeff Sebo, expect the probability to rise over time. The question is still open.

Are my feelings for an AI companion real?

Yes. Your feelings happen in you, and they are real. The open question is whether the AI experiences anything, and current science suggests it does not.