Animal Cognitive Uplift

Speculative If the step from a measured change in an animal’s brain to a claim about that animal’s experience can be closed — and nothing in this field has closed it — the prize is a causal account of which particular genetic changes built the human brain, stated in neurons and cortical folds rather than in correlations across genomes. The condition comes first because the whole subject rests on that gap. Comparative genomics can rank candidate human-specific changes; it cannot say which does anything, and the only test anyone has proposed is to put one into a living animal and look.

Established The magnitude is best stated in the units the experiments themselves report. One human-specific gene, ARHGAP11B, expressed at physiological levels in fetal common marmoset neocortex, produced an approximately three-fold increase in mitotic basal radial glia in the outer subventricular zone, near 40% more Satb2-positive and near 50% more Brn2-positive upper-layer neurons, and a gyrification index that moved from about 1.0 to about 1.1 where folds appeared at all. Human MCPH1 copies in rhesus macaques delayed the peak of cortical grey-matter volume by an average of 164 days and improved delayed-matching-to-sample performance at every delay tested, the weakest of those comparisons at P = 0.0086. Three hundred thousand human glial progenitor cells injected into newborn mice left a forebrain progenitor population that was largely or entirely human within twelve months. No comparative-genomics study returns a number of that kind about a living brain.

Frontier The medical return sits at the cell-therapy end of the field, which is not the end that draws attention. The human glial progenitors that colonise a host forebrain are the same cells being developed for remyelination; the human cortical tissue that integrates with a host cortex is the same material being developed for repair after stroke and injury. Nothing clinical depends on a macaque with better working memory. The gains from the transgenic-primate branch accrue to a short list of laboratories and to the evolutionary question itself; the costs fall on animals that cannot be asked.

Handwave The condition on all of it is a stopping rule, and nobody has one. The laboratory holding the strongest anatomical result destroyed its transgenic marmoset fetuses 50 days before term because it could not predict postnatal brain function, and said so in print. That is a decision, not a rule: it binds one group, in one country, for one experiment. 11 sets out why informed people disagree about what the rule should be; 7 names the study that would settle the empirical half, and the reason it may be the study that ought not to be run.

Established Four different experimental programmes get filed under one headline, and they are not the same experiment. The first introduces a gene found in humans and no other living species into another animal and asks what its brain does differently. The second replaces an animal’s copy of a shared gene with the human version — a knock-in. The third puts living human cells into an animal brain, a cellular chimera. The fourth grafts laboratory-grown human brain tissue into an animal brain. Only the first two alter the genome; only the last two put human cells inside it; and an ethical argument that lands against one frequently does not touch the others.

Established “Uplift” is a term from science fiction and no laboratory in this field uses it. The 2021 National Academies committee on this subject went out of its way to recommend against the loaded vocabulary that had grown up around the work, naming “mini-brain” as inaccurate and “humanized” as implying a boundary has been crossed, and preferring “neural organoid”, “neural transplant” and “neural chimera”. This brief keeps the commissioned title because that is the phrase the public argument is conducted under, and examines it in 15.

Frontier The distinction the whole subject turns on is between a change in an animal’s brain and a change in an animal’s mind, and every measured result to date is on the first side of it. A gyrification index, a neuron count and a latency in a touchscreen task are measurements of a body. Whether any of them corresponds to a difference in what it is like to be that animal is an inference, and the instruments that would test the inference are the subject of an unsettled research programme of their own, treated in Consciousness Research.

Established This brief deliberately withholds method. It is written to be usable by a legislator or a research ethics committee and useless as a protocol. Constructs, delivery systems, embryo handling and germline procedures are named where the argument needs them to exist, and are not described.

Established The scale of the human material involved is small, and that fact does real work in the ethical argument. Laboratory-grown human neural organoids reach roughly 4 mm across and carry on the order of 100,000 neurons against an estimated 86 billion in an adult human brain. The 2021 National Academies committee concluded that transplants of human neural cells into non-human animals “currently involve far too few cells to generate any capacity approaching consciousness”. The counter-argument, which 4 takes seriously, is that cell count is the wrong axis: the marmoset result changed no cell’s species, and the glial chimeras changed almost every glial cell’s species and no neuron’s.

Established ARHGAP11B is a human-specific gene, and its effect on a primate neocortex has been measured once. The gene arose by partial duplication of ARHGAP11A on the human lineage roughly five million years ago, and a single later base substitution created a 47-amino-acid sequence unique to humans and necessary for its effect on brain growth. Earlier work in 2015 put it into mouse and ferret embryos at levels the authors themselves call unphysiological, where it expanded the neocortex and folded brains that are normally smooth. The 2020 experiment is the one that matters, because it used a primate and physiological expression.

Established Seven transgenic common marmoset fetuses were obtained; five expressed the gene in fetal neocortex; all were analysed at embryonic day 101 of a gestation of about 150 days. The neocortex was significantly longer than in controls (P < 0.05) though not wider; the gyrification index rose from about 1.0 to about 1.1 where gyrus-like structures appeared; mitotic basal radial glia in the outer subventricular zone increased about three-fold (P < 0.001); and Satb2-positive and Brn2-positive upper-layer neurons — the classes that expand most in primate evolution — rose by nearly 40% and nearly 50% (P < 0.001). The work was done by the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden with the Central Institute for Experimental Animals in Kawasaki and Keio University in Tokyo, and published in Science in June 2020.

Established The fetuses were not carried to term, and the reason given is explicitly ethical rather than technical. The paper states: “We confined our analyses to marmoset fetuses because we anticipated that expression of this human-specific gene would affect neocortex development in this animal. In light of potential unforeseeable consequences with regard to postnatal brain function, we considered it a prerequisite — and mandatory from an ethical point of view — to first determine the effects of ARHGAP11B expression on the development of fetal marmoset neocortex.” The same sentence appears, lightly edited, in the institutional press releases. 11 examines what it does and does not commit anyone to.

Frontier The marmoset result says nothing about cognition, and its authors do not claim otherwise. No behaviour was measured, because there was no behaving animal, and the inference from more upper-layer neurons to more capacity is a hypothesis rather than a finding. Three expert commentaries on the paper — from Pawel Stankiewicz, Zoltan Molnar and David Rowitch — treat it as a decisive demonstration of the gene’s role in neocortical expansion, and none raises an ethical or interpretive objection.

Frontier The one experiment that carried transgenic primates to adulthood and tested them behaviourally is the human MCPH1 macaque study, published in National Science Review in 2019 by a group at the Kunming Institute of Zoology. Eleven transgenic rhesus monkeys were produced — eight first-generation, three second-generation — carrying two to nine copies of human MCPH1. Of the eight, one died 76 days after birth, two were lost to abortion at embryonic day 136, and five survived to testing; all three second-generation animals were euthanised as embryos. The procedures were approved in advance by the institutional animal care and use committee of the Kunming Institute of Zoology and the Yunnan Key Laboratory of Primate Biomedical Research.

Frontier The measured phenotype is a developmental delay, not an enlargement. Longitudinal imaging put the peak of cortical grey-matter volume on average 164 days later than in controls, which the authors read as human-like neoteny of brain development, and transcriptome analysis found 970 differentially expressed genes at embryonic day 136 and 1,933 at postnatal day 76, with differentiation and maturation genes suppressed. In a delayed-matching-to-sample task they outperformed controls at every delay reported — P = 0.0086 at 0 to 4 seconds, 0.0022 at 8 seconds, 0.00055 at 16 seconds, 0.032 at 32 seconds, 0.00075 combined — and had shorter reaction times.

Frontier The brain-size claim attached to this study in secondary coverage is contested by the study’s own text. The primary paper reports that absolute brain volume was initially smaller in the transgenic monkeys, that the difference closed by about three years, and that relative brain volume showed no significant difference over development. A companion piece by two of the same authors states instead that the transgenic monkeys had “similar brain volume and cortex thickness as the controls, but their relative brain volume and gray matter percentage were higher”. Wire coverage at the time reported flatly that brain size did not increase. This brief does not resolve the discrepancy; it records that the headline number of the field’s only behavioural transgenic-primate experiment is reported differently in the paper and in its authors’ own commentary.

Frontier The MCPH1 result has not been replicated, and it rests on five animals in one laboratory. Five survivors, tested against a control group in one facility, on a battery not registered in advance, with several delay intervals each generating a comparison. Nothing about that design is unusual for primate neuroscience, and nothing about it supports the weight the public argument places on it. The absence of an independent replication seven years on is the most informative fact about it; 5 says why one is unlikely to arrive.

Established The cellular-chimera branch has the field’s cleanest engraftment numbers. Three hundred thousand human glial progenitor cells delivered to newborn mice across five sites had, within twelve months, produced a forebrain glial progenitor population “largely, and often entirely, of human origin”; in three of four animals at one year no mouse glial progenitors were found in the sampled sections. Replacement ran corpus callosum by three months, superficial cortical layers by eight, essentially complete by twelve.

Frontier The behavioural claim attached to those animals is the field’s second anchor, and this brief could not open the primary paper. Han and colleagues reported in Cell Stem Cell in 2013 that mice engrafted as neonates with human glial progenitors showed as adults enhanced long-term potentiation and improved performance across several learning tasks. The publisher refused retrieval this session, so the content here comes from independent descriptions: the 2014 engraftment paper from the same group, summarising the result as “enhanced long-term potentiation and improved performance in a variety of learning tasks”, and the 2021 National Academies report, recording that the adult hosts “exhibited improved performance in learning and memory tasks compared with controls”. The direction and existence of the effect are well attested. The effect sizes, the task list and the authors’ caveats are not here, because this brief did not read them.

Frontier Organoid transplantation is the newest branch and the one where host welfare has actually been measured, in both directions. Human brain organoids grafted into adult mouse brain in 2018 were vascularised by host vessels and became synaptically interconnected with host tissue; the same report is cited for host mice performing worse in spatial memory after surgery, a de-enhancement that belongs in the record beside the gains. A 2022 transplantation into the somatosensory cortex of newborn rats produced grafts occupying roughly a third of a hemisphere, neurons more mature than the same tissue achieves in culture, graft responses to stimulation of the host’s whiskers, and reward-seeking behaviour driven by optogenetic stimulation of graft neurons — with no other difference from controls reported. Later work integrated human forebrain organoids with the injured adult rat visual system and traced axons along corticospinal tracts in mice and cynomolgus monkeys.

Speculative The FOXP2 knock-in mouse is the oldest example in this literature and the one this brief can support least. The claim in circulation is that mice carrying the two human-specific amino-acid substitutions in Foxp2 show altered ultrasonic vocalisation, changed exploratory behaviour and altered striatal plasticity, and that a later study found faster transitions between learning strategies in a maze. No source this brief could retrieve mentions FOXP2 at all: the primary journals refused and the search budget was spent. The claims are stated because they are what a policy reader will meet, and flagged at a level this brief can defend rather than the level the literature probably supports.

Established What has not been shown is stated as plainly in the field’s own consensus document as anywhere else. The 2021 National Academies committee concluded that there is “currently no objective basis for ascribing consciousness to organoids”, that current models are “extremely unlikely” to possess awareness, emotion or pain in the foreseeable future, and that assessing either requires behavioural observation of a whole animal rather than measurement of neural activity. It recorded no demonstration, in any system, of enhanced cognition beyond specific task contexts, of altered emotional capacity in a host, or of human-like experience in an animal carrying human neural cells. The measured record is a record of bodies.

Established One further result sets the outer edge of what has been done. A team of United States, Spanish and Chinese researchers produced the first human-monkey chimera at the Kunming Institute of Zoology in July 2019: human pluripotent cells contributed to a primate embryo, not carried to term. 10 says which regimes single that class out for prohibition rather than review.

Frontier The first open question is whether a count of progenitors licenses any claim about capacity, and it is open in both directions. The marmoset experiment measured basal radial glia and upper-layer neurons because their abundance tracks neocortical expansion across primates. That correlation is the strongest reason to care about the result and the weakest possible support for a claim about what the animal could have done. Nobody has run the study that would connect them, for the reason 7 gives.

Frontier The second is whether the macaque working-memory effect is real. Five animals, one facility, no pre-registration, multiple delay intervals each yielding a test, and a phenotype — delayed maturation — that could affect how an animal engages with an apparatus independently of what it remembers. The effect may be exactly what the authors say; the design cannot distinguish that from a modest false positive, and only replication can.

Frontier The third is the discrepancy inside the MCPH1 record itself. A paper reporting no significant difference in relative brain volume and a companion note from two of its authors reporting higher relative brain volume and grey-matter percentage cannot both describe the same imaging dataset correctly. Resolving it requires access to the data and a stated analysis plan, and nobody has published the resolution. A field whose single behavioural primate result carries an unreconciled headline number cannot yet tell a legislature what it has found.

Frontier The fourth is whether the glial chimeras are enhanced at all, or merely running human astrocyte physiology inside a mouse circuit. Human astroglia are morphologically distinct from mouse astroglia, and a mouse forebrain whose glial progenitors are entirely human is less a mouse with extra ability than a mouse with a different cell type doing a familiar job. If the learning improvement is species-typical for human glia, there is no threshold being approached, only a substitution; if it is an enhancement, the threshold question is live. No published experiment discriminates the two.

Speculative The fifth is whether there is any experiment that could detect a morally relevant change in experience, as opposed to a change in performance. The National Academies committee holds that consciousness assessment depends on observing a whole behaving animal, so the question can only be asked of an animal already created and raised. Alexandre Erler has proposed operational markers — mirror self-recognition, and false-belief tests of the kind used with great apes — as thresholds that could in principle be applied. Whether passing them would show anything about experience rather than about cognitive sophistication is the dispute Consciousness Research and Machine Consciousness record as unresolved.

Frontier What would discriminate, across all five, is a small number of comparisons nobody has published. An equal-volume graft of non-human tissue against a human graft, on the same battery, separates “human cells” from “more cells”. A blinded, pre-registered replication in a second facility separates the effect from the laboratory. A welfare instrument built for a novel phenotype rather than a known disease model separates distress from difference. None requires a technique that does not exist.

Established The binding constraint on the transgenic-primate branch is not technique but arithmetic. Eleven transgenic macaques yielded five testable adults, in a programme whose imaging arm measured a 164-day shift in a developmental peak and therefore ran for years. A replication starts from the same losses, the same gestation, the same maturation and the same matched control colony. No statistical fix closes that gap; it sets how many transgenic primate cognition experiments the world can run at all, which is a small number per decade.

Frontier There is no validated cognitive battery designed to detect a novel capacity, as opposed to a deficit. Primate cognitive testing is built around disease models and lesion studies, where the expected direction is loss and the comparison is a species-typical baseline. A battery distinguishing “better at this task” from “capable of something this species does not do” is a different instrument, and nobody has built one. That is why the macaque result is reported as a higher score and a shorter reaction time rather than as a new ability.

Frontier There is no welfare instrument for an animal whose phenotype nobody has seen before. The 2021 National Academies committee warned that animals in this class “are likely to show behaviors that resemble human symptoms and that would be viewed as distressing were they to occur in humans”, and recommended close observation. Against what reference? Welfare scoring in primate facilities is calibrated to species-typical behaviour; an animal that departs from its species norm by design defeats the calibration, and the departure is the experiment. It is the most concrete unsolved problem in the field, and the one an ethics committee meets first.

Established The organoid-transplantation branch is constrained by immune status and vascular supply. Grafts are typically placed in immunodeficient hosts to prevent rejection, so the behaving animal is not a normal animal and carries that confound into every comparison. Graft size is capped by perfusion: human neural tissue stalls below roughly half a millimetre without a blood supply, and the 2018 result that host vessels invade a graft is what made larger transplants possible. The vascularisation ceiling is treated in Lab-Grown Organs and Whole Organ Regeneration.

Frontier Expression level and mosaicism decide whether an experiment means anything, and are hard to control. An animal in which only some cells carry the introduced gene, or carry it at the wrong level, gives a phenotype that is neither the human condition nor a clean control. The MCPH1 animals carried between two and nine copies, a fourfold spread inside a cohort of five tested survivors.

Frontier The literature has a direction problem that no methodological fix addresses. An experiment that finds nothing in a transgenic primate is expensive, unpublishable where a positive result would have gone, and awkward for a programme that spent years of animal life on it. The field has published no null from a transgenic-primate cognition study, which under these incentives is what a field with few positive results and no replications looks like either way. 7 proposes registration as the only available remedy.

Speculative The last bottleneck is legal. A study carried to adulthood in a non-human primate needs a facility, a regulator and a funder simultaneously willing; 10 shows the overlap is narrow and moving in opposite directions in different jurisdictions, which describes regulatory arbitrage rather than a bottleneck that gets solved.

Frontier This subject waits on results held by neighbours, none of which is producing them for this purpose. It does not wait on a missing technology — every intervention in 3 has been performed — but on the ability to interpret what it has already done.

Frontier It depends first on the science of detecting experience in a system that cannot report it. Consciousness Research records the field’s largest adversarial test damaging both of its leading theories; Machine Consciousness records that a verdict on any given system is fixed by the theory the assessor brought with them. Both transfer: with no theory-neutral test for experience in a machine there is none for a chimera, and every threshold proposal in 11 inherits that.

Frontier It depends second on comparative cognition having benchmarks that survive being applied across species. Interspecies Communication records the same difficulty elsewhere: detecting structure in a species’ signals is not showing that the structure carries meaning. A battery showing a transgenic macaque scoring above a control macaque has shown a difference between two macaques and nothing about where either sits on a scale that includes humans.

Frontier It depends third on the genetics of human cognition being better resolved than it is. Neurogenetics records how much of a polygenic education signal survives a within-family comparison, which decides whether a single-gene intervention in an animal models anything human. A human-specific gene is not thereby a gene that explains a human capacity.

Speculative It depends fourth on the developmental biology of tempo, because the only measured transgenic-primate phenotype in the field is a tempo phenotype. Developmental Tempo Control owns whether human developmental slowness is causal or incidental, and the MCPH1 result — a 164-day delay in a grey-matter peak, read by its authors as human-like neoteny — is a claim in that currency. If tempo is a consequence rather than a driver, the reading of the only behavioural primate experiment in this field changes without any new animal being made.

Frontier It depends fifth on the tissue side, where Organoid Intelligence and Embryo Models hold the evidence. The first owns what cultured human neurons can be shown to do and how a sentience claim about them would be tested; the second owns the oversight categories a three-dimensional human construct is sorted into.

Frontier The decisive experiment is a pre-registered, blinded, adequately powered behavioural and welfare study of a non-human primate carrying a human neurodevelopmental variant and carried to adulthood, with welfare as a primary endpoint rather than a monitoring condition. It is decisive because every disputed claim in this brief turns on one missing link: whether a measured change in a primate brain corresponds to any change in what that primate can do, and at what cost.

Handwave The honest answer may be that the decisive experiment should not be performed, and this brief does not treat that as a rhetorical gesture. Two independent arguments reach it. The harm-benefit argument: no clinical programme depends on the result, the beneficiary is a question in evolutionary biology, and the subject is a long-lived social primate held for years in conditions chosen for the experiment. The moral-status argument: the experiment is designed to succeed, and success would produce an animal whose standing nobody agreed in advance how to assess, in a facility with no procedure for an ambiguous assessment. A brief whose decisive experiment is ethically foreclosed has found something, and this one reports it rather than routing around it.

Established A partial version of the decisive experiment has already been performed once, which is the strongest single fact in this brief. The MCPH1 macaques were carried to adulthood and behaviourally tested; five of eleven animals survived to testing; the battery was not pre-registered; welfare was not a reported endpoint; and no independent group has repeated it in the seven years since. The experiment is not hypothetical and not prospective. It is unreplicated.

Established The laboratory holding the strongest anatomical result declined the same experiment in advance, in print, on ethical grounds. That refusal is itself a policy result, and it is why this brief calls the decisive experiment unscheduled rather than forthcoming. It is also a precedent with no enforcement: one group’s stated prerequisite binds no other group in any other jurisdiction.

Frontier The second experiment on the list, and the cheapest, is an independent blinded replication of the macaque working-memory battery. It needs no new technique, no new gene and no ethical category beyond the one crossed in 2019, and would establish whether the field’s only behavioural transgenic-primate result survives a second laboratory. Its absence is not explained by difficulty.

Frontier The third is a species-control graft, and it is the experiment this brief would most like to see run because it is nearly free. Transplant an equal volume of non-human neural tissue against a human graft into matched hosts and run the same battery. If only the human graft changes behaviour, the claim that human cells do something particular is supported for the first time. If both do, much of the public argument collapses into a question about tissue volume. No published experiment reports the comparison.

Frontier The fourth is a welfare-primary study of existing chimeric and transgenic animals, using instruments built for a novel phenotype rather than a disease model. It addresses the National Academies’ own warning about behaviours that would read as distress in a human, and runs on animals that already exist, which makes it the only item here that creates no new subject.

Speculative The fifth is the analysis nobody has published: the reconciliation of the MCPH1 imaging discrepancy. A stated analysis plan applied to the existing scans would settle whether relative brain volume differed, at the cost of one dataset and one preprint.

Speculative The sixth is institutional: a public registry of neural chimera and transgenic-neurodevelopment protocols, with pre-registered endpoints and mandatory reporting of nulls. No mechanism exists by which a legislature could learn that such an experiment was attempted and failed.

Speculative The seventh is a natural experiment already running, and it is the only one on this list with no start date to negotiate. Jurisdictions are diverging, as 10 sets out. Watching where the work goes over a decade measures how much any of these instruments constrains, and nobody has to approve it.

Handwave What no experiment here can do is settle whether any of these animals has an inner life, and the brief says so rather than leaving it implied. The consensus position is that consciousness assessment requires observing a whole behaving animal; that animal does not exist; and the instruments are contested independently of this subject. An experiment that cannot in principle answer the question the public argument is about is still worth running, and worth saying that it would not end the argument.

Established The engineering requirements for the decisive experiment are conventional, expensive and slow, and none of them is a research problem. A breeding colony of a primate species with multi-year maturation, housed to psychological-wellbeing standards; longitudinal structural imaging across the whole developmental window, because the only measured phenotype in the field is a shift in when a curve peaks; cognitive rigs that run unchanged for years; and staff continuity across a decade. The 164-day figure at the centre of this brief is a difference between two developmental trajectories, and a trajectory cannot be measured in a short grant.

Established The chimera branch needs less, which is part of why it has produced more. Barrier housing for immunodeficient rodents, neonatal surgery, and cohorts held for the twelve months on which a human glial population replaces a mouse one. That is within reach of an ordinary university department, and the capital difference between the branches predicts the shape of the literature better than any ethical argument does.

Frontier The instrument that does not exist is the welfare rig. Continuous behavioural monitoring calibrated to an animal with no species-typical reference, with pre-agreed intervention thresholds and an observer who does not report to the principal investigator. Every component exists; the combination is standard nowhere, and no funder requires it.

Frontier On the tissue side the binding engineering problem is perfusion, and it is shared. Human neural tissue cannot be grown past roughly half a millimetre without a blood supply, so the 2022 result that a graft could occupy about a third of a rat hemisphere depended on the host’s vasculature rather than on any improvement in culture. Lab-Grown Organs and Biological Robots carry the manufacturing side of this constraint; Organoid Intelligence carries the measurement side.

Frontier The nearest technology is the one that removes the animal. Organoid Intelligence asks what cultured human neurons can be made to do without a body around them. If it succeeds it removes the welfare objection from a subset of these questions and replaces it with the one that brief owns: whether a dish of human neurons can be a subject.

Frontier The nearest method is cell therapy, and it shares every technique and almost none of the controversy. Human glial progenitors that remyelinate a host brain and human cortical tissue that integrates with a host cortex are the same preparations under a different intent; Lab-Grown Organs and Whole Organ Regeneration carry the clinical side. Intent is what separates a chimera experiment from a preclinical study, and intent is not a property an oversight body can measure.

Speculative The nearest question is the human one. Cognitive Enhancement is the brief this one is most often confused with, and its central finding is the discipline needed here: every honest meta-analysis of making a healthy adult think better returns an effect between 0.1 and 0.3 standard deviations. A field that cannot reliably improve cognition in humans, where consent is available and the instruments are validated, should not have its animal branch described in terms of large gains. Biological Enhancement and Human Cognitive Augmentation carry the rest of that comparison.

Speculative The nearest conceptual neighbour is the argument about where cognition starts. Basal Cognition records good measurements carried by a contested word, which is this brief’s structure with the species reversed; Cognitive Liberty holds the rights vocabulary any future claim on behalf of such an animal would borrow, since none exists for the animal itself.

Established There is no international instrument governing this work, and the national ones were written for adjacent problems. What exists is animal-research law, human-tissue law, stem-cell guidance and funding conditions: a patchwork, in the 2021 National Academies committee’s word, addressing specific concerns rather than a new field.

Established In the United States the principal instrument is a funding condition never finalised as policy. In 2015 the National Institutes of Health imposed a moratorium on funding research introducing human pluripotent stem cells into pre-gastrulation embryos of any non-human vertebrate; replacement oversight proposed in 2016 was never finalised, and the moratorium stayed in effect. The 2009 NIH guidelines separately prohibit funding research that breeds any animal where human stem cells may contribute to its germline. Neither reaches privately funded work.

Established The 2021 National Academies report is the closest thing the field has to a consensus document, and its own interests are worth stating. It was produced under contract to the National Institutes of Health with support from The Dana Foundation, co-chaired by Bernard Lo and Joshua Sanes, with a committee including Paola Arlotta, R. Alta Charo, Fred Gage, Henry Greely and Patricia King. Several members lead laboratories whose programmes the report evaluates — a structural interest rather than an accusation, since a committee of the only people who understand a field is a committee of people with something at stake.

Established Its recommendation is a three-tier scheme, and the tiers are specific. Proceeding under existing mechanisms: neural organoids from patient-derived cells, transplants into wild-type mice, in vitro chimeras. Requiring additional review: large-scale primate neural transplants, organoids with ambiguous donor consent, mouse blastocyst chimeras carried to implantation. Not to proceed at present: human stem cells into primate blastocysts with implantation, primate gamete generation, and research violating donor consent. The committee found existing oversight under the Animal Welfare Act and Public Health Service policy well developed for transplantation work, while noting that institutional committees may lack independent expertise for this class of protocol.

Established The International Society for Stem Cell Research is the field’s own body and its guidance is not law anywhere. The ISSCR recommends specialised review where research could produce a high degree of integration into an animal’s central nervous system or human gametes in non-human hosts, and especially rigorous review for non-human primates; its 2021 guidelines place current human brain organoid work in the category exempt from specialised oversight. Julian Koplin’s 2023 response is the sharpest published criticism, and its claim is narrow and hard to dismiss: the recommendations fail to address the core ethical concern raised by neurological chimeras, which is moral status rather than integration.

Established The United Kingdom prohibits by guidance the experiment this brief calls decisive. Home Office guidance adopted in 2016 sorts animals containing human material into three categories: national specialist review is required for substantial modification of an animal’s brain that may make its function potentially more human-like, particularly in large animals, and prohibited experiments include transplanting enough human-derived neural cells into a non-human primate to make substantial functional modification of its brain possible. Beneath it sits the Animals (Scientific Procedures) Act 1986, requiring a harm-benefit analysis by the Secretary of State.

Frontier The same jurisdiction barely regulates the tissue side, the gap a 2025 analysis in the Medical Law Review identifies. Because cell lines derived from cultured cells are not relevant material under the Human Tissue Act 2004, and the Human Fertilisation and Embryology Act reaches the work only where human embryonic stem cells are derived inside the project, brain organoid research from banked induced pluripotent cells needs no research ethics committee approval. Emily Jackson proposes a dedicated oversight committee and developmental limits borrowed from embryo research. Nothing has been enacted.

Established Germany, Japan and China diverge, and the divergence is the institutional fact that matters most. The German National Ethics Council adopted a three-tiered approach holding that transferring human cells into mammals other than primates is ethically acceptable under conditions and that great-ape research should be prohibited; a separate analysis notes that extending welfare protection to laboratory-grown human brain tissue in Germany might need a constitutional amendment, since the research is protected by constitutional scientific freedom that only a comparable interest can limit. In March 2019 the Japanese science ministry issued guidance permitting human induced pluripotent cells to be injected into non-human animal embryos for basic research; it neither permits nor prohibits non-human primates, and great-ape research is not allowed in Japan. Oversight of neural transplants and human neural chimeras in China is described by the National Academies report as unclear; both the MCPH1 macaques and the 2019 human-monkey embryo chimera came from the same institute, under an institutional animal care committee.

Frontier The most recent institutional movement is regional rather than national. An Asia Pacific neuroethics working group convened in Singapore in November 2024 and published what it calls the region’s first comprehensive ethical framework for human brain organoid research: adaptive regulation, consent covering future uses, and explicit attention to the welfare and potential enhancement of chimeric animals, noting that the moral weight attached to blurring the human-animal boundary varies across cultures. It is guidance, not law.

Frontier Interested parties, named. Researchers running transgenic-primate or chimera programmes have careers and facilities depending on the work continuing, and a reason to present welfare questions as tractable. Funders have a reason to present the clinical route as shorter than it is. Animal-protection organisations have a reason to treat any measured difference as a morally relevant change — the same inferential leap the field is criticised for making in the other direction. Scientific societies issuing their own guidelines are at once the regulated party and the drafter. A reader should discount in all four directions, and the Institute does not say which discount is larger.

Established The facts that do not depend on anyone’s position come first. Seven marmoset fetuses were produced and destroyed at embryonic day 101 of a 150-day gestation. Of eleven transgenic macaques, one died at 76 days, two were lost at embryonic day 136, three were euthanised as embryos, and five reached behavioural testing. Rodent hosts of organoid grafts are typically immunodeficient, and one published report describes host mice performing worse on spatial memory after surgery. None of these animals could consent, none was a patient, and the benefit accrued to someone else. Those are the facts, neither worse nor better than they sound.

Frontier The strongest form of the “this is premature” position is Insoo Hyun’s, and it is not a dismissal. On this view a change in moral status requires distinctively human capacities — propositional belief and recursive thought of the kind tied to language mastery — which arise not from tissue but from conditions of human child-rearing that no laboratory animal encounters. Concern about moral humanisation is then not merely early but mistargeted, and the tractable questions are scientific justification, incremental experimentation and animal welfare, which existing oversight already asks. The position is defensible, restrictive in practice, and held by someone closely associated with the guidelines that implement it.

Frontier The strongest opposing position is Alexandre Erler’s, and it attacks the threshold rather than the timing. Requiring uniquely human capacities sets the bar where most humans’ moral standing does not sit. Erler adopts David DeGrazia’s account of great apes as borderline persons — mirror self-recognition, planning, tool use, grasp of social hierarchy, without propositional language — and notes that animal-research law already gives great apes protections others do not get. If that threshold is defensible for apes, it is defensible for an animal engineered toward it. He concludes that debate is urgent rather than premature, that mirror and false-belief tests could serve as operational markers, and that an animal accidentally reaching full moral status should not then be killed.

Frontier A third position abandons humanness as the wrong axis. Chen and colleagues argued in 2019 for replacing moral humanisation with enhancement, since an enhanced trait need not be uniquely human and welfare is owed whether or not the animal resembles a person. A 2023 analysis in that spirit anticipated rodent hosts of human grafts showing changes in sensory sensitivity, motor function, some memory and prosocial behaviour, while judging higher-level capacities much less likely. The frame is answerable — it becomes a welfare-science question 7 can name an experiment for — and it concedes the question most of the public argument is about.

Speculative A fourth position points at the transplanted human material rather than at the animal. Andrea Lavazza has argued that transplanting human neural tissue capable of sentience into an animal violates its dignity. Tsutomu Sawai and colleagues develop this as a dual-status problem: a graft experiment can hold two entities with moral standing, and the tissue’s standing may exceed its host’s, making otherwise permissible animal research impermissible because of what is inside the animal. Their conclusion is that no immediate change is needed and the interval should be used to prepare — the commonest position in the ethics literature and the least satisfying to either side.

Frontier The fifth position is that existing frameworks are adequate, and it carries institutional weight. The National Academies committee found that transplantation research falls under a well-developed oversight system, that most of the work should proceed under existing mechanisms, and that the near-term risk of consciousness is remote. Against it: harm-benefit analysis, the engine of every animal-research regime in the world, presupposes the subject’s moral status is fixed and known before the study starts — the one assumption this class of experiment disturbs. Both can be true; the system may be well developed and still be answering a different question.

Frontier The sixth position is the abolitionist one, and it is usually summarised unfairly. Its strongest form does not claim a chimeric animal has human standing; it claims the harm-benefit test fails on its own terms when the benefit is a question in evolutionary biology, the subject is a long-lived social primate, and the cost is years of its life. Jacqueline Glover’s objection to the MCPH1 work had this shape: to humanise them is to cause harm — where would they live and what would they do?

Frontier The most striking disagreement in the record is internal to one of the studies. Martin Styner, a co-author on the MCPH1 paper, said afterward that in his personal opinion such research should not be done. Bing Su, the senior author, held that an ethics board had approved it and no humanness was observed, only one gene having changed. Megan Dennis and Anthony Chan defended it on evolutionary distance and translational value; Arthur Caplan objected that inserting human genes into monkey brains is a different matter from other transgenic work; Larry Baum noted that the study altered a few positions in one of roughly 20,000 genes. These are not misunderstandings of the evidence but different weightings of it, and the Institute does not adjudicate between them.

Established Public opinion is neither the alarm nor the permission that either side reports it to be. Survey work cited by the National Academies found 52% of Americans opposed to the use of animals in research, with more opposition to chimeric animals than others. A 2024 programme of three studies found almost no perceived blurring of the boundary: asked about rats, pigs and chimpanzees carrying human brain tissue implants, 76.7%, 75.2% and 73.3% of respondents rated the animal as zero percent human, 84.5% placed its moral rights at the scale midpoint, and agreement that such implants blur the line averaged 2.01 on a seven-point scale. The public is more opposed to the practice, and less confused about the category, than either side assumes.

Handwave What nobody has is a stopping rule. Not a principle — there are several — but a rule naming an observation and saying what happens when it is made: which committee is told, on what timescale, and what it may do about an animal that already exists. The German Ethics Council, the UK Home Office and the National Academies all draw their lines at the level of experiment type; none specifies a procedure triggered by a result. That absence is the field’s central governance fact, and why the marmoset laboratory’s refusal in 3 does work no institution is doing.

Speculative The durable consequence is not a smarter animal; it is a legal category under load. Every jurisdiction sorts entities into persons and property, and animal-welfare law works inside the second category by limiting what may be done to property. An animal engineered toward capacities the law treats as grounds for heightened protection in great apes presses on that sorting from a direction it was not built to take. No case has arisen, and the instruments that would decide one presume a species-typical subject.

Frontier The second consequence is jurisdictional and it is already visible. One state prohibits substantial functional modification of a primate brain by guidance; one permits blastocyst work by 2019 ministry guidance silent on primates; one has a funding condition that never became policy and does not reach private money; one is described by the field’s own consensus report as unclear. Work migrates to where it is permitted, and the migration is the measurement — not of which regime is right but of which regimes bind. Nobody is collecting the data.

Speculative The third is precedent that runs the other way. Techniques for putting a human gene into a primate embryo are the techniques for putting an edited gene into a human embryo, and the governance conversation about the second runs separately from the first. Genetic Engineering and Embryo Models carry that thread; the animal work is where the methods mature and the oversight is thinnest.

Frontier The fourth is reputational, and a research council will feel it first. A field that reports a result as making monkeys smarter, then turns out to have measured a delay in a grey-matter curve and a reaction time, spends credit it will need later. The committee’s recommendation against the words mini-brain and humanized was governance rather than style: 52% opposition to animal research is a standing political fact, and terminology that oversells is the cheapest way to convert it into policy.

Handwave The fifth is irreversibility, stated without drama. An animal that has been made cannot be unmade, and the decision to make it is taken by a laboratory and an institutional committee, ahead of any public process, on a harm-benefit judgement about an animal whose harms are not yet characterised. That is not an argument against the work but a description of who decides, and it holds in every jurisdiction named in 10.

These horizons track what the evidence and the governance could plausibly look like, not what anyone has scheduled. The decisive experiment in 7 has no date because no one has proposed it.

• 10 yr: Frontier More chimera and graft results, more precisely measured, with host welfare reported as an endpoint in a minority; probably no independent replication of the macaque working-memory result; possibly a species-control graft comparison, which is cheap and would be the decade’s most informative publication. Regulatory divergence widens rather than converges.
• 25 yr: Speculative Either a second transgenic-primate cognition programme has run and been replicated, making the threshold argument in 11 an applied question with data; or it has not, and the field is one unreplicated result plus a large tissue-transplantation literature. A welfare instrument built for novel phenotypes is the plausible arrival in this window.
• 50 yr: Speculative If any animal has been shown on agreed instruments to exceed its species-typical range on a capacity the law already treats as grounds for heightened protection, the legal category question in 12 becomes a live case rather than a hypothetical. Nothing about the current trajectory makes that likely, and nothing about it rules it out.
• 100 / 250+ yr: Handwave Claims about engineered non-human persons belong here and nowhere earlier. They require a theory of experience the field does not have, a stopping rule nobody has written, and an instrument nobody has built. They are stated in this corpus so that they can be labelled, which is the only honest thing available to do with them.

• Depends on Nothing on this map blocks the experiments themselves; all of them have been performed. What is blocked is interpretation, and the evidence sits with neighbours. Consciousness Research and Machine Consciousness own the question of whether any test for experience is theory-neutral, which every threshold proposal here inherits. Neurogenetics owns whether a single human-specific gene models a human capacity. Developmental Tempo Control owns whether developmental slowness is causal, which decides how to read this field’s only behavioural primate result. Embryo Models owns the oversight categories a three-dimensional human construct is sorted into.
• Requires (not on this map) A cognitive battery that tells a higher score apart from a capacity the species does not have, since every existing primate instrument was built to detect loss; a welfare instrument for an animal whose species-typical baseline the experiment deliberately removed; a volume-matched non-human graft arm, without which no behavioural chimera result separates human cells from more cells; a stopping rule that names an observation, says who is told and says what they may do, since every existing line is drawn at the level of experiment type; a registry that makes a failed transgenic-primate study visible to a legislature; and colony capacity able to carry a matched control arm across a primate’s maturation, the constraint that sets how many such experiments the world can run per decade.
• Enables Causal attribution of specific neuroanatomical features to specific human-specific genetic changes, which comparative genomics cannot supply; human glial and cortical grafts as therapeutic preparations for demyelination and cortical injury; a tested rather than assumed answer to whether measured neural change tracks change in capacity; and, if the welfare instrument arrives, a harm-benefit analysis that works on a subject whose status is not fixed in advance.
• Adjacent Organoid Intelligence is the branch that removes the animal and inherits a different subject question; Interspecies Communication shares the problem of reading capacity off structure; Biological Robots and Lab-Grown Organs share the tissue-manufacturing ceiling; De-Extinction Technologies shares the rhetorical failure mode of reporting edits as outcomes; Cognitive Enhancement supplies the effect sizes against which any claim of large cognitive gain should be read.

Handwave “Uplift” imports its conclusion. The word presumes a measured change is an improvement along a scale running toward humans, and nothing in the record establishes a scale, a direction or an improvement. A marmoset fetus with more upper-layer neurons has more upper-layer neurons.

Speculative “They are building Planet of the Apes.” The defence offered at the time — that rhesus monkeys are far enough from humans that a few added genes will not produce that scenario, and that the study touched a few positions in one of about 20,000 genes — is correct as stated and does not answer the question being asked, which is about thresholds rather than totals. Alarm and reassurance both treat the number of genes as the relevant quantity, and nobody has shown that it is.

Frontier “The monkeys got smarter.” What was measured was a 164-day delay in the peak of cortical grey-matter volume, a higher score at four delay intervals in five animals, and shorter reaction times. Absolute brain volume was initially smaller than in controls and converged by about three years, and whether relative brain volume differed at all is reported differently by the paper and by its authors’ commentary. No general cognitive measure was taken, because none exists for this purpose.

Established “They created human-brained monkeys and they are alive somewhere.” The marmosets carrying the human-specific gene were never born; all analysis was at embryonic day 101 of a 150-day gestation, and the paper says the decision was ethical rather than technical. The macaques carrying human MCPH1 copies were born and five reached testing — but MCPH1 is not human-specific, it is a shared gene present in humans in a different form.

Speculative “FOXP2 is the language gene, so humanising it gave mice a piece of language.” The selective-sweep evidence that made FOXP2 famous was later challenged in the same literature that produced it, and the gene is now usually described as one input to vocal-motor circuitry rather than a gene for language. This brief could retrieve neither the knock-in paper nor the reanalysis and states both at low confidence; the reason is in 3. What is not in doubt is that the mouse work measured vocalisation, exploratory behaviour and striatal plasticity, and that none of those is language.

Established “Brain organoids are mini-brains.” The 2021 National Academies committee recommended against the term as inaccurate. Organoids reach roughly 4 mm and on the order of 100,000 neurons against an estimated 86 billion in an adult human brain, lack the multi-region circuitry the committee treats as a precondition for consciousness, and supply no objective basis for ascribing it.

Frontier “Human cells in an animal brain make it partly human.” The claim is untested in the only form that would settle it: no published experiment compares a human graft against a volume-matched non-human graft on the same battery, so nothing in the literature distinguishes an effect of human cells from an effect of additional cells. The 2024 survey work suggests the public does not make the inference anyway — three quarters of respondents rated animals carrying human brain tissue implants as zero percent human.

Established “This is unregulated.” It is regulated unevenly, by instruments written for other problems: a United States funding moratorium since 2015 and a germline-breeding prohibition since 2009; national specialist review and an outright prohibition on substantially modifying a primate brain in United Kingdom guidance; a German Ethics Council tiering that would prohibit great-ape work; permissive Japanese guidance from 2019 silent on primates; and an oversight picture in China the National Academies report calls unclear. There is no international instrument specific to this work, and the most widely cited document, the ISSCR guidance, comes from the field’s own professional society and is law nowhere. As of August 2024 no country had a framework specific to human brain organoid research, and what is closest to unregulated is organoid work from banked induced pluripotent cell lines in the United Kingdom, which needs no research ethics committee approval.

Frontier “Ethics committees already answer this.” They answer harm-benefit, which requires the subject’s moral status to be known before the study begins — the assumption this class of experiment disturbs, and the gap Koplin identifies in the ISSCR settlement and Erler in the three-Rs framework. A committee can be competent, diligent and correct within its remit and still not be the body that answers the question.

Speculative “The science is decades away, so the ethics can wait.” The experiment that would force the question was performed in 2019, and the refusal that currently substitutes for a rule was recorded in 2020. Erler’s reply to the prematurity claim is that debate is urgent precisely because the empirical question is being decided one protocol at a time, and this brief finds that description accurate whatever one concludes about the threshold.

Handwave “Either these animals suffer more or they do not; someone must know which.” Nobody knows which. The National Academies committee expects animals in this class to show behaviours that would read as distress in a human, and recommends close observation without saying against what reference. It is the field’s largest answerable question, and 7 names the study that would answer it.