Home / The Big Question

Can AI Be Conscious? What the Science Says Now

A glowing node in a web of living filaments above a night city

Can AI be conscious? For the AI systems that exist today, the most detailed scientific assessment so far says no: a 2023 report by 19 researchers concluded that no current AI system is conscious. The same report found no obvious technical barrier to building systems that meet the conditions its theories describe, so whether a future machine could be conscious remains open.

That split answer, unlikely now but not ruled out, is where most careful researchers stand. The rest of this guide explains how they got there and why they still disagree.

How researchers ask whether AI can be conscious

There is no agreed test that detects consciousness from the outside, not even in animals. Researchers therefore work from theories. Each major theory of consciousness says which kind of processing it thinks matters, and that description can be checked against how an AI system is built.

The 2023 report, written by philosophers, neuroscientists and AI researchers including Patrick Butlin, Robert Long, Yoshua Bengio and Jonathan Birch, did exactly that. It turned several leading theories into a list of "indicator properties" stated in computational terms, then looked for those properties in recent AI systems. Its verdict on the systems it examined was negative, and it added that nothing obvious stands in the way of building systems that have the indicators. The guide on how to test if AI is conscious walks through that checklist.

Five glowing drifting forms in a line over a beach

Why today's chatbots are judged unlikely to be conscious

The philosopher David Chalmers looked at large language models in particular. He argued that they are probably not conscious, because mainstream theories point to features these models lack: recurrent processing, a global workspace and unified agency. He also argued that these gaps could plausibly be closed within about a decade, and that successors to today's models may well be conscious.

Global workspace theory, the source of one of those features, is explained in our guide to global workspace theory and AI. What the models themselves say about their inner lives is a separate question, covered in is ChatGPT conscious.

An hourglass in a desert filled with drifting motes of light

Why the question is still open

Two things keep the debate alive.

First, the theories disagree about what consciousness needs. Some are framed in terms of what a system does, so a machine that does the right things would count. Others tie consciousness to the physical structure doing the work. Researchers developing Integrated Information Theory argued in 2024 that a computer could match a conscious system's behaviour exactly without sharing its experience. On that view, functional equivalence does not imply phenomenal equivalence, and copying what a brain does would not be enough. Our guide to integrated information theory and AI explains why.

Second, expert opinion is divided rather than settled. In a 2020 survey reported by 80,000 Hours, nearly 40% of philosophers leaned towards future AI systems being conscious, against about 27% leaning against. AI developers have taken note: when Anthropic announced a research programme on model welfare in April 2025, it wrote that there is no scientific consensus on whether current or future AI systems could be conscious.

A closer look: "not now" is not "never"

It is easy to read the 2023 report's conclusion as a final answer. It is a statement about the systems the authors examined, judged against theories that may themselves be wrong. The same document that says no current system is conscious also says the barriers to building one are not obvious. That is why researchers who doubt today's systems still argue for preparing now, and why the question has moved from science fiction into research programmes and lab policies.

The words matter too. "Conscious" and "sentient" are often used as if they meant the same thing, and they do not; the guide to AI sentience vs consciousness sets out the difference.

What to keep in mind when you read a claim about conscious AI

  • Which theory is it using? A claim that a system is or is not conscious only makes sense relative to some account of what consciousness is.
  • Is it about today's systems or future ones? Most careful verdicts are confident about the first and cautious about the second.
  • Does it rest on what the AI says? A fluent answer about feelings is not evidence of feelings; language models are trained on human writing about exactly that.
  • Who is making it, and on what evidence? A peer-reviewed assessment and a viral conversation screenshot are not the same kind of evidence.

More in The Big Question · Report a correction