Can brain organoids become conscious? What science can and cannot tell us

Whether brain organoids can become conscious is an open scientific and philosophical question, but there is no evidence that any organoid made so far is conscious. Theories of consciousness disagree about what would be required, no validated test exists for tissue without behavior, and a 2021 National Academies report judged consciousness in today’s organoids extremely unlikely.

Key takeaways

• Established: no brain organoid has shown evidence of consciousness. Some organoid activity resembles features of preterm infant EEG (Trujillo et al., 2019), but resemblance in a recording is not evidence of experience.
• Broad consensus: current organoids lack most of what plausibly matters for consciousness: scale, blood supply, natural sensory input, a body, and the long-range organization of a brain.
• Open dispute: the leading theories (integrated information, global workspace, recurrent processing, higher-order) give different answers about what an organoid would need, and none is settled (Seth & Bayne, 2022).
• Hard problem in practice: the best consciousness markers, such as the perturbational complexity index, were calibrated on human brains and cannot simply be transferred to a dish.
• The live ethical question is not “are today’s organoids conscious?” but “what precautions are proportionate as assemblies, vascularization, transplants and closed-loop embodiment make organoids more complex?”

Why are people asking whether brain organoids can be conscious? #

A brain organoid is a three-dimensional cluster of neural tissue grown from stem cells, usually induced pluripotent stem cells made from donated skin or blood. Our guide to how brain organoids are grown explains the process. Organoids self-organize into rough versions of developing brain regions and produce spontaneous electrical activity. That alone would not make headlines. Two results did.

Organoid activity that resembles preterm EEG #

Trujillo et al. reported in Cell Stem Cell (2019) that cortical organoids cultured for months developed increasingly complex oscillatory network activity, recorded on microelectrode arrays. The authors compared features of this activity with EEG recordings from preterm infants and found that a model trained on the infant data could estimate the developmental age of the organoids from their activity.

This is a genuine and useful finding about how human neural networks develop. It is often reported as “organoids show brain waves like a premature baby.” That framing invites a leap the data do not support. The comparison was between statistical features of electrical signals, not between experiences. Preterm infants are themselves a case where the onset of consciousness is debated, and an organoid shares only a narrow slice of a preterm brain’s properties: it has no thalamus wired to a body, no sensory organs and no blood supply.

The DishBrain “sentience” wording #

Kagan et al. reported in Neuron (2022) that cultured neurons on a high-density microelectrode array learned to improve at a simplified game of Pong when given structured sensory feedback in a closed loop. The paper’s title said the neurons “exhibit sentience.” The authors used the term in a specific technical sense: being responsive to sensory input in a way that changes behavior, framed by the free energy principle.

Balci et al. responded in the same journal (2023), arguing that applying words like “sentience” and “intelligence” to this result was inappropriate and risked misleading the public and other scientists. The disagreement was partly about vocabulary and partly about whether the result showed goal-directed learning at all. Either way, the episode showed how a technical term can be read as a claim about feelings. Our article on how neurons learn in a dish covers the experiment itself.

Together, these two results explain why the question keeps returning whenever organoid intelligence is in the news. Neither shows consciousness. Both show that organoid systems are becoming more brain-like in some measurable respects, and that the language used to describe them matters.

What does “conscious” actually mean here? #

Much confusion in public debate comes from people using one word for several different things. Philosophers and neuroscientists draw distinctions that are worth keeping straight.

Phenomenal versus access consciousness #

Thomas Nagel’s classic paper (1974) framed the core idea: a system is conscious if there is something it is like to be that system. This is what Ned Block (1995) called phenomenal consciousness, the felt, subjective quality of experience. Block distinguished it from access consciousness: information being available for reasoning, report and the control of behavior.

The distinction matters for organoids. Access consciousness is defined by what a system can do with information, and organoids can do very little: they cannot report, reason or act on their own. Phenomenal consciousness is defined by experience itself, which is exactly what we cannot observe directly in any system other than ourselves. When people ask whether an organoid could be conscious, they almost always mean the phenomenal sense, and that is the hardest sense to test.

Sentience versus sapience #

Two further terms often get blurred:

• Sentience usually means the capacity for experiences with a positive or negative quality, such as pain, pleasure or distress. This is the sense most relevant to welfare and moral status.
• Sapience refers to higher cognitive capacities: reasoning, planning, self-awareness, language.
• Intelligence, in the machine learning sense, is about adaptive problem solving. It can exist without either sentience or sapience. A thermostat-like controller or a trained neural network can adapt without anyone thinking it feels anything.
• Wakefulness or arousal is a physiological state (awake versus asleep). It is linked to consciousness in humans but is not the same thing.

These can come apart. A system could learn a task (some intelligence) with no experience at all. It is also conceivable, on some theories, that a system could have simple experiences with no intelligence worth the name. For ethics, sentience is the property that matters most, because it is the basis for asking whether something can be harmed. The question of what follows morally, if anything, is the subject of our companion piece on the moral status of biological computers.

What do the main theories of consciousness predict for brain organoids? #

There is no agreed scientific theory of consciousness. Seth and Bayne’s review in Nature Reviews Neuroscience (2022) grouped the major contenders into families, including higher-order theories, global workspace theories, theories built on re-entry and predictive processing, and integrated information theory. They noted that these theories often aim to explain different aspects of consciousness and make different empirical commitments.

Recent work has started testing theories against each other. The Cogitate Consortium reported in Nature (2025) a large adversarial collaboration that pre-registered competing predictions from integrated information theory and global neuronal workspace theory and tested them with several brain imaging methods. The results supported some predictions of each and challenged key claims of both. In other words, even in adult humans, the field has not converged.

Because each theory identifies different ingredients, each gives a different verdict on organoids.

Three patterns stand out from this comparison.

1. Almost every theory gives current organoids a very low probability. The exception is IIT, which does not rule out a minimal degree of experience in any sufficiently integrated physical system. Critics see this as a weakness of IIT; supporters see it as a consequence of taking experience seriously.
2. The theories point to different warning signs. IIT points to connectivity and integration. Workspace theories point to multi-region architecture. Embodied views point to sensory feedback loops. A lab that only watches one of these may miss what another theory would treat as significant.
3. Theoretical uncertainty is itself a reason for care. If we do not know which theory is right, we cannot simply pick the most convenient one. This is the core argument for precaution, discussed below.

Croxford and Bayne, writing in Neuroethics (2024), offer a clear version of the skeptical position. They argue from two broad constraints on any adequate theory, that consciousness depends on embodiment and on representation, that disembodied neural organoids are not plausible candidates for consciousness, while not disputing that caution is needed for some organoid types. That qualification matters: the case against is strongest for isolated organoids and weaker for embodied, connected or transplanted ones.

How could you test a brain organoid for consciousness? #

In humans, clinicians judge consciousness mainly from behavior and report: can the patient follow a command, answer a question, track an object? Organoids cannot do any of that, so behavioral tests are useless. The search has turned to brain-based markers that do not depend on behavior.

The perturbational complexity index #

The most prominent candidate is the perturbational complexity index (PCI). The idea grew out of work by Massimini et al. (2005) in Science, which found that a magnetic pulse delivered to the cortex produced widespread, complex responses during wakefulness but brief, local responses during deep sleep. Casali et al. turned this into a number in Science Translational Medicine (2013): stimulate the cortex with transcranial magnetic stimulation, record the EEG response, and measure how complex and non-redundant the resulting pattern is. The index distinguished conscious from unconscious states across wakefulness, sleep, anesthesia and brain injury in the people they studied.

Casarotto et al. reported in Annals of Neurology (2016) an empirical cutoff derived from a benchmark population and used it to stratify unresponsive patients, identifying some who appeared vegetative at the bedside but had high complexity. PCI is attractive for organoids because it is explicitly independent of sensory processing and behavior, and because it was inspired by the integration and differentiation ideas behind IIT.

Lavazza and Massimini argued in the Journal of Medical Ethics (2018) that techniques like PCI, already used with brain-injured patients who cannot communicate, could in principle be applied to cerebral organoids, and that if organoids showed signs of sensibility, an ethical discussion of their use would become necessary.

Why these markers are hard to apply to organoids #

The difficulty is calibration. Every validated marker of consciousness was validated on human brains, where we had an independent check: people could report their experience after waking or recovering. Organoids offer no such check.

• Different scale. An organoid has far fewer neurons than a brain. The human brain has about 86 billion neurons (Azevedo et al., 2009); many organoids have fewer than 100,000 cells and are often under 500 micrometers across (Smirnova et al., 2023). A complexity threshold set for a whole cortex has no obvious meaning for a speck of tissue.
• Different stimulation and recording. PCI uses magnetic pulses through the skull and scalp EEG. In an organoid the analogues are electrical stimulation and recording on electrodes, a related but different measurement. See our explainer on microelectrode arrays for how these work.
• The circularity problem. To validate a marker in a new kind of system, you need to know whether that system is conscious. But that is exactly the question the marker was meant to answer.
• False positives and negatives. Complex activity can arise in systems no one thinks are conscious, and some theories hold that consciousness could exist with signals our instruments do not capture.
• Activity is not experience. The National Academies report (2021) made this point directly: measurements of neural activity alone are not sufficient to establish consciousness or pain.

The fair conclusion is that markers like PCI are useful as screening signals, not verdicts. A sustained rise in perturbational complexity in an organoid system would not prove experience, but it would be a reasonable trigger for closer review.

What do current brain organoids lack? #

Whatever theory one prefers, current organoids fall short on nearly every dimension that theories treat as relevant. The comparison below focuses on those dimensions; our ethics overview has a broader table of organoid versus brain properties.

• Size and cell number. Organoids are tiny, and their networks are small and local.
• Vasculature. Without blood vessels, oxygen and nutrients reach the interior only by diffusion, so larger organoids tend to develop a starving core. This caps their growth. See organoid vascularization.
• Sensory input. Organoids receive no light, sound or touch. Any input is electrical or chemical stimulation chosen by researchers.
• Motor output and a body. They control nothing in the world unless coupled to software.
• Large-scale organization. They lack the thalamocortical loops, layered cortex and long-range connections between many regions that workspace and recurrent theories emphasize. Their spontaneous activity often takes the form of network bursts, synchronized waves rather than differentiated, sustained patterns.
• Maturity. Gordon et al. reported in Nature Neuroscience (2021) that long-term cortical organoid cultures progress through transitions resembling early postnatal development, which is a significant advance but still far from a mature brain.

These limitations are why the scientific consensus, to the extent one exists, judges consciousness in current organoids very unlikely. They are also a list of what researchers are actively trying to improve, for good scientific and medical reasons.

What could change the picture? #

The question is evergreen because the technology is moving. Each of the following developments addresses one of the gaps above.

Assembloids #

Assembloids join organoids representing different brain regions. Birey et al. reported in Nature (2017) that fusing dorsal and ventral forebrain spheroids allowed interneurons to migrate and integrate into cortical circuits. Linking regions is a step toward the multi-region architecture that workspace theories care about, although current assembloids remain small and simple.

Vascularization #

Cakir et al. reported in Nature Methods (2019) a way to engineer brain organoids with a vascular-like system. Better perfusion could allow larger, longer-lived and more complex organoids, removing one of the main physical caps on their growth.

Transplantation into animals #

Mansour et al. reported in Nature Biotechnology (2018) that human brain organoids transplanted into mouse brains became vascularized by the host. Revah et al. reported in Nature (2022) that human cortical organoids transplanted into newborn rats matured, received sensory and other inputs from the host brain, produced sensory responses, and could influence reward-seeking behavior when stimulated. Transplants solve the vasculature and sensory input problems at once, which is why ethics bodies treat them as a more sensitive category than organoids in a dish. Biocomputing on electrode arrays does not require transplantation.

Embodiment through closed loops #

Biocomputing gives organoids something they never have in ordinary culture: structured input and feedback. Closed-loop stimulation, as in DishBrain, or input encoding in reservoir computing with brain organoids, couples living tissue to a virtual environment. This is still a very thin form of embodiment: a few electrodes, a simple task, no pain receptors. But embodied and predictive theories would regard sustained, rich sensorimotor loops as the most relevant direction to monitor, which places biocomputing research squarely inside the debate.

None of these developments has produced evidence of consciousness. The point is that the conditions most theories consider relevant are no longer fixed, so a one-time answer will not stay current. Our piece on the future of biocomputing covers where the technology is heading.

Precaution or overcaution: how should uncertainty be handled? #

When we cannot resolve whether a being can suffer, there are two ways to be wrong. We can treat a system that has experiences as if it had none, or we can impose heavy costs to protect a system that has no experiences at all. Reasonable people weigh these errors differently.

The case for precaution #

The precautionary view holds that the stakes of the first error are severe, so uncertainty should push us toward protection. Koplin and Savulescu, writing in the Journal of Law, Medicine & Ethics (2019), argued that the possibility of consciousness could set moral limits on some kinds of brain organoid research and discussed how to proceed under uncertainty. Jonathan Birch’s The Edge of Sentience (2024) develops a general framework for “sentience candidates,” systems for which there is a realistic possibility of sentience, and argues for precautions proportionate to that possibility, decided through inclusive and accountable processes. Neural organoids are among the cases the book discusses. Farahany et al. (2018) called in Nature for ethical guidance to be developed before brain surrogates become more sophisticated, not after.

The case against overcaution #

The opposing concern is that precaution without evidence can block valuable work. Organoids are used to study neurodevelopmental disorders, test drugs on human tissue and reduce some animal experiments. Restricting them based on a possibility that no current theory strongly supports would have real costs, borne largely by patients. Skeptics also note that if mere possibility is enough to trigger protections, the principle could expand without limit, since almost any neural tissue might be conscious on some theory.

Where the positions meet #

In practice, the positions converge more than headlines suggest. Few serious voices call for a halt on current organoid research; few deny that more complex systems warrant more scrutiny. The disagreement is over thresholds: which developments should trigger extra review, and what that review should involve. The decision of when, if ever, it becomes wrong to end an organoid culture is covered in our article on the ethics of switching off biological computers.

What did the National Academies conclude? #

The most detailed institutional assessment to date is the 2021 report of the US National Academies of Sciences, Engineering, and Medicine, The Emerging Field of Human Neural Organoids, Transplants, and Chimeras. Its conclusions on consciousness, paraphrased, were:

1. Given their limited complexity, current neural organoids are very far from having awareness, emotions or the capacity to experience pain, and this was judged extremely unlikely in the foreseeable future.
2. Organoids, at present, do not warrant greater moral standing than other human neural tissue studied in the lab.
3. Measuring neural activity alone is not enough to determine whether consciousness or pain is present, and most existing assessment methods depend on behavior, which organoids lack.
4. There is no agreed definition or measure of consciousness that could settle the question for organoids.
5. Existing oversight was judged generally adequate for current research, but the question should be revisited as organoids, transplants and chimeras become more sophisticated.

Two features of the report are often lost in coverage. First, it was explicitly time-bound: it judged the state of the science at the time, and it anticipated the need for reassessment. Second, it separated organoids in a dish from transplants into animals, treating the latter as raising more pressing questions. Anyone citing the report should keep both points.

What should responsible researchers measure and disclose? #

Because there is no test for consciousness, the practical goal is not to certify organoids as non-conscious. It is to make the relevant facts visible so that concerns can be checked rather than argued about in the abstract. A reasonable reporting practice for any lab, including biocomputing groups, would include the following.

Labs can also set internal review triggers in advance, for example a sustained, unexplained rise in complexity measures, the introduction of transplantation, or long-duration embodiment in rich environments. Committing to triggers before results arrive is a guard against motivated reasoning in either direction. Consent matters here too: donors whose cells become organoids may care about these uses, which connects to questions explored in who owns a living computer.

What remains genuinely open? #

Some questions are empirical and may be answered with better tools. Others are philosophical and may never be fully settled. Keeping them distinct helps.

Empirical questions:

• How do complexity and integration measures change as organoids mature, fuse or become vascularized?
• Can perturbation-based markers be meaningfully adapted to small tissue on electrode arrays, and what would a calibration even look like?
• Does closed-loop embodiment change network organization in ways relevant to any theory?

Theoretical questions:

• Which theory of consciousness, if any, is correct? Adversarial collaborations are beginning to test this in humans.
• Is consciousness graded, so that tiny amounts are possible, or does it switch on above some threshold of organization?
• Is a body, or at least a sensorimotor loop, necessary?

Ethical questions:

• How much probability of sentience is enough to trigger protections, and who decides?
• How should the interests of possible but unproven experience be weighed against concrete benefits to patients?

The bottom line #

Can brain organoids become conscious? The established answer is that no organoid has shown evidence of consciousness, and the broad scientific view is that today’s organoids almost certainly are not. The honest answer is that we cannot prove this from first principles, because science lacks an agreed theory and a validated test for systems without behavior. And the forward-looking answer is that the gap is not fixed: assembloids, vascularization, transplantation and closed-loop embodiment each narrow it in ways some theories would care about.

That combination calls for neither alarm nor dismissal. It calls for measurement, transparent reporting, pre-agreed review points and precise language. At Organoid Programming we are building provenance, open data and experiment records into our platform from the start, because questions like this one are easier to answer well when the evidence is recorded and shared.

Frequently asked questions

Are brain organoids conscious?

There is no evidence that any brain organoid made so far is conscious. A 2021 National Academies report judged that awareness, emotion or pain in organoids is extremely unlikely in the foreseeable future, because they are small, lack blood vessels, receive no natural sensory input and lack the large-scale organization of a brain. The report also said the question should be revisited as organoids become more complex.

Could a brain organoid ever become conscious in the future?

Nobody can rule it out, and nobody can confirm it. Whether it is possible depends on which theory of consciousness is right, and that is still disputed. Most theories tie consciousness to large-scale integration, recurrent loops between regions, or a body interacting with the world. Assembloids, vascularized organoids, transplantation and closed-loop embodiment each move organoids a little closer to some of those conditions, which is why researchers favor ongoing monitoring.

What is the difference between consciousness, sentience and intelligence?

Consciousness, in the sense philosophers call phenomenal consciousness, means there is something it is like to be a system: it has experiences. Sentience usually means the capacity for experiences that feel good or bad, such as pain or pleasure. Intelligence, or sapience, is about solving problems and adapting behavior. These can come apart: a system could learn a task without experiencing anything, which is why task performance alone says little about consciousness.

Did the DishBrain study show that neurons in a dish are sentient?

Kagan et al. (2022) used the word sentience in a narrow technical sense: neurons that adjust their activity in response to sensory feedback, framed by the free energy principle. They did not claim the culture had feelings. Balci et al. (2023) argued in the same journal that this wording risked misleading readers. The experiment showed closed-loop adaptation, not evidence of experience.

Can the perturbational complexity index test organoids for consciousness?

Not yet in any validated way. The perturbational complexity index measures how complex and integrated the brain’s response to a magnetic pulse is, and it was calibrated on human brains during waking, sleep, anesthesia and brain injury. Lavazza and Massimini (2018) suggested adapting it to organoids, but an organoid is a different system, so there is no agreed threshold. A high score would be a reason for closer scrutiny, not proof.

Which theory of consciousness is most likely to say an organoid could be conscious?

Integrated information theory is the most permissive, because it treats consciousness as graded and tied to the causal structure of a physical system rather than to behavior or reports. On that view, a richly interconnected organoid could have some small degree of consciousness. Global workspace and higher-order theories set the bar much higher, requiring broadcast across specialized brain systems or self-monitoring representations that current organoids do not appear to have.

Should brain organoid research be stopped until we know for sure?

Most ethicists and scientific bodies do not argue for a halt, since there is no evidence of consciousness in current organoids and the research has medical value. The common position is proportionate precaution: monitor activity, set review checkpoints for more complex models such as assembloids or transplants, and report findings openly. Critics of overcaution note that blocking research also has costs, including slower progress on brain disease.

What should labs measure to address organoid consciousness concerns?

Labs can report organoid size, age, cell composition and vascularization, record spontaneous and evoked activity on microelectrode arrays, describe any sensory feedback or closed-loop training, and track measures of network complexity and integration over time. Sharing raw data in open formats lets others check for unexpected changes. None of these measures proves or disproves consciousness, but together they make concerns testable.

Sources

1. Trujillo et al. (2019). Complex oscillatory waves emerging from cortical organoids model early human brain network development. Cell Stem Cell.
2. Kagan et al. (2022). In vitro neurons learn and exhibit sentience when embodied in a simulated game-world. Neuron.
3. Balci et al. (2023). A response to claims of emergent intelligence and sentience in a dish. Neuron.
4. Nagel (1974). What is it like to be a bat? The Philosophical Review.
5. Block (1995). On a confusion about a function of consciousness. Behavioral and Brain Sciences.
6. Seth & Bayne (2022). Theories of consciousness. Nature Reviews Neuroscience.
7. Tononi, Boly, Massimini & Koch (2016). Integrated information theory: from consciousness to its physical substrate. Nature Reviews Neuroscience.
8. Dehaene & Changeux (2011). Experimental and theoretical approaches to conscious processing. Neuron.
9. Lamme (2006). Towards a true neural stance on consciousness. Trends in Cognitive Sciences.
10. Brown, Lau & LeDoux (2019). Understanding the higher-order approach to consciousness. Trends in Cognitive Sciences.
11. Cogitate Consortium et al. (2025). Adversarial testing of global neuronal workspace and integrated information theories of consciousness. Nature.
12. Massimini et al. (2005). Breakdown of cortical effective connectivity during sleep. Science.
13. Casali et al. (2013). A theoretically based index of consciousness independent of sensory processing and behavior. Science Translational Medicine.
14. Casarotto et al. (2016). Stratification of unresponsive patients by an independently validated index of brain complexity. Annals of Neurology.
15. Lavazza & Massimini (2018). Cerebral organoids: ethical issues and consciousness assessment. Journal of Medical Ethics.
16. Croxford & Bayne (2024). The case against organoid consciousness. Neuroethics.
17. National Academies of Sciences, Engineering, and Medicine (2021). The Emerging Field of Human Neural Organoids, Transplants, and Chimeras: Science, Ethics, and Governance.
18. Smirnova et al. (2023). Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. Frontiers in Science.
19. Azevedo et al. (2009). Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled-up primate brain. Journal of Comparative Neurology.
20. Gordon et al. (2021). Long-term maturation of human cortical organoids matches key early postnatal transitions. Nature Neuroscience.
21. Birey et al. (2017). Assembly of functionally integrated human forebrain spheroids. Nature.
22. Cakir et al. (2019). Engineering of human brain organoids with a functional vascular-like system. Nature Methods.
23. Mansour et al. (2018). An in vivo model of functional and vascularized human brain organoids. Nature Biotechnology.
24. Revah et al. (2022). Maturation and circuit integration of transplanted human cortical organoids. Nature.
25. Birch (2024). The Edge of Sentience: Risk and Precaution in Humans, Other Animals, and AI. Oxford University Press.
26. Koplin & Savulescu (2019). Moral limits of brain organoid research. Journal of Law, Medicine & Ethics.
27. Farahany et al. (2018). The ethics of experimenting with human brain tissue. Nature.