De-Extinction Technologies

Established Three quite different technical claims share the word de-extinction, and conflating them is the single most common error in the subject. Back-breeding selects among living descendants for an ancestral appearance. Cloning copies a genome from preserved cells of an animal that actually lived. Genome editing rewrites a living relative's genome toward a target phenotype inferred from ancient DNA. Established Only the second has ever produced a living animal of an extinct taxon: a Pyrenean ibex cloned in 2003 from cells of Celia, the last of her subspecies, who had died in 2000. Established The clone died shortly after birth of a lung defect. That is the complete record of extinct animals restored to life, and it lasted minutes.

Established The case that made the subject famous is the third kind, and it is weaker than its coverage. In April 2025 Colossal Biosciences announced three dire wolves — Romulus, Remus and Khaleesi — produced by making twenty edits across fourteen genes in grey wolf endothelial progenitor cells. Established No ancient dire wolf DNA was inserted into the genome. Established Colossal's own Chief Scientific Officer, Beth Shapiro, subsequently described the animals as grey wolves with twenty edits; the IUCN Species Survival Commission's Canid Specialist Group stated they are not a dire wolf under any definition and that the project does not contribute to conservation.

Established Meanwhile the thing that worked was a ferret. Elizabeth Ann, born 10 December 2020, was cloned from cells of Willa, a black-footed ferret who died in the mid-1980s leaving no living descendants; her genome carries three times as much genetic diversity as any modern black-footed ferret, and clones from the same line produced surviving kits in June 2024. Established It required no editing, no new species concept and no revised definition of extinction — only a cryobank, somatic cell nuclear transfer, and a studbook. Frontier The gap between those two stories is the subject of this brief, and it is not mainly a gap about genomics.

Established Back-breeding is a hundred-year-old programme with a settled verdict, and the verdict is about genotypes. Heinz and Lutz Heck began selectively breeding cattle for an aurochs phenotype at the Hellabrunn and Berlin zoos in the early 1920s; the Union of German Zoo Directors condemned the effort in 1951, resolving to give no financial or other support for breeding back extinct species. Established Heck cattle were released at Oostvaardersplassen in the Netherlands in the 1980s as aurochs surrogates, large numbers died in the harsh winters of 2005 and 2010, and the hands-off management regime ended in 2018. Established Later programmes are technically better and make the same claim: Taurus cattle from 1996 crossed Heck stock with Sayaguesa, Chianina and Spanish fighting bulls, and the Tauros Programme works with Rewilding Europe on a twenty-year horizon toward self-sufficient herds of more than 150 animals, with a Scottish Highlands reintroduction scheduled for 2026. Established The decisive sentence in the aurochs literature is that approaches aiming at breeding an aurochs-like phenotype do not equate to an aurochs-like genotype, and the general finding on back-breeding is that it produces phenotypic look-alikes often lacking in behavioural, physiological and ecological traits of the original species.

Established The most intellectually honest programme in the field solved this by renaming its product. Reinhold Rau began the Quagga Project in 1987 with nineteen plains zebras from Namibia and South Africa selected for reduced rear striping; the first foal arrived in 1988, and by early 2006 third- and fourth-generation animals were judged visually similar to museum specimens. Established The project calls its animals Rau quaggas and states plainly that they will resemble the quaggas only in external appearance but will be genetically different. Frontier That single naming decision does more epistemic work than any of the technical advances of the intervening forty years, because it converts an unfalsifiable identity claim into a falsifiable appearance claim. Speculative A field-wide convention requiring proxies to be named after their maker rather than their target would cost nothing and would end most of the disputes in this brief.

Established Cloning is the route with a real success, and its numbers are the numbers that govern everything else. Dolly required 277 reconstructed eggs to yield 29 viable embryos, three lambs at birth and one surviving to adulthood; the same experiment is also quoted as 435 attempts on a different denominator, and the two framings are not the same statistic and should not be swapped. Established Prometea, the first cloned horse, took 814 attempts. Established In one cattle cloning experiment a third of calves died young. Established The number that matters here is that somatic cell nuclear transfer success rates in wild animals remain under one per cent: a cloned gaur died after two days, a banteng survived. Established Success in domestic species has improved sharply — reported rates of seven to eight in ten by 2014, and Sooam Biotech producing 500 cloned embryos a day by 2016 — but that improvement rests on abundant oocytes and decades of embryo-transfer infrastructure, and neither exists for elephants, thylacines or pigeons.

Established Elizabeth Ann is the result the field should be pointing at, and it is the one it does not lead with. She was the first cloned North American endangered species, produced from cells that had sat in the San Diego Zoo's Frozen Zoo since the mid-1980s, in a partnership between the zoo, Revive & Restore, ViaGen Pets and Equine and the US Fish and Wildlife Service. Established She had a hysterectomy for hydrometra in October 2022 — a condition common in the species and not attributed to cloning — and died in 2025 without reproducing. Established Two further clones from Willa's line, Noreen and Antonia, were born in April 2024; in June 2024 Antonia produced three kits, two of which survived, with further litters in 2025. Established A bottlenecked population descended from a handful of founders now has, in its breeding pool, the genome of an animal that died before most of its living members' grandparents were born. Frontier That is a measurable increase in a real species' genetic diversity, achieved with cryopreservation and cloning alone, and it is the strongest evidence in the subject that the technology has conservation value.

Established The editing route has raised the most money and produced the least defensible claim. Colossal has raised $435 million since founding: $15 million seed in September 2021, a $60 million Series A in March 2022, a $150 million Series B in January 2023 at a valuation over $1 billion, $50 million for the Colossal Foundation in October 2024, and a $200 million Series C in January 2025 at a $10.2 billion valuation, with Form Bio spun out in September 2022 on $30 million. Established The dire wolf announcement was accompanied by a scientific paper, and the paper's most important property is what it does not contain. Frontier Gedman and colleagues, posted to bioRxiv on 11 April 2025, is a phylogenomics preprint: two dire wolf paleogenomes at 3.4× and 12.8× coverage from individuals dating to more than 13,000 and more than 72,000 years ago, consensus species trees across those and ten other canid species, an ancestry model in which roughly two-thirds of dire wolf ancestry derives from a lineage sister to the grey wolf, coyote and dhole clade and roughly one-third from near the base of Canini, and eighty genes identified as evolving under diversifying selection. Established It makes no de-extinction claim, it does not present the edits, it is not peer reviewed, and George R. R. Martin is a listed author alongside George Church and Ben Lamm.

Established The peer-reviewed prior points the same direction by a different route, and neither primary supports the premise of the engineering. Perri and colleagues, in Nature in 2021, sequenced five dire wolf genomes from sub-fossil remains between 13,000 and more than 50,000 years old and found a highly divergent lineage that split from living canids around 5.7 million years ago, with no evidence for gene flow between dire wolves and either North American grey wolves or coyotes. Frontier The two papers disagree about hybridisation — deep isolation against a two-thirds/one-third hybrid origin — and agree that the dire wolf sits outside Canis. Established On the 2025 preprint's own shortlist of what distinguishes a dire wolf, eighty genes are candidates and fourteen were edited, and the two sets are not stated to be the same genes. Established For scale, a diagnostic human clinical genome is typically sequenced at about 30×; the reference against which the edits were chosen rests on two animals, one of them at coverage low enough that most heterozygous sites cannot be resolved. Frontier A press release and a preprint were issued on the same day about different things, and only one of them was peer-adjacent.

Established The mammoth programme's public evidence is a mouse with unusual hair. The March 2025 preprint from Chen and colleagues reports optimised workflows achieving simultaneous editing of up to seven different genes, including loss-of-function mutations in Fgf5, Tgm3 and Fam83g, and describes the resulting mice as displaying curly, textured coats and golden-brown hair. Established It reports no cold-tolerance measurement, and positions itself as a rapid platform for testing mammoth-centric genetic variants — a modest and defensible claim. Established Robin Lovell-Badge of the Francis Crick Institute questioned whether any mechanism was elucidated and whether cold tolerance had been demonstrated at all; Victoria Herridge noted that engineering mammoth-like elephants is a categorically harder problem than engineering mice. Established The distance to be closed is arithmetic: woolly mammoth and African elephant genomes are 98.55 to 99.40 per cent identical, which across a roughly 3.2-gigabase genome is on the order of twenty to forty-five million differences. Established Mammoth genome sequencing was completed in 2015 and mammoth variants were edited into Asian elephant cells the same year, and the standing assessment of all three revival routes remains that none of these approaches are currently feasible.

Established The right benchmark for a mammoth calf by 2028 is a rhinoceros programme that has everything Colossal lacks and still has no birth. Two northern white rhinos remain, Najin and Fatu, both female, at Ol Pejeta Conservancy in Kenya and owned by Dvur Kralove Zoo. Established BioRescue has produced 38 cryopreserved embryos as of 2025, stored in Berlin and Cremona, using frozen sperm from dead males and oocytes from a living female of the target subspecies, with induced pluripotent stem cell work underway and historical DNA from a dozen individuals banked in Berlin and San Diego. Established In 2023 a southern white rhino embryo transfer succeeded, and the surrogate, Curra, died of a bacterial infection in November. Frontier That is assisted reproduction with living cells of the target taxon, a closely related surrogate species with established embryo-transfer protocols, and more than a decade of funded effort — and no live birth. Speculative Elephant reproductive technology is substantially less developed than rhinoceros reproductive technology, which is why a stated 2028 timeline for mammoth calves should be read as a fundraising horizon rather than a scientific projection.

Established The institutional position was set before the announcements and has not moved. The IUCN Species Survival Commission established a De-extinction Task Force in April 2014 to develop principles governing the creation of extinct-species proxies for conservation purposes, and issued guiding principles in 2016 — the standard against which the dire wolf was judged and found wanting. Established The Canid Specialist Group's objection was not only that the animals are not dire wolves but that phenotypic proxies do not change an extinct species' conservation status, and that the effort may threaten extant grey wolves. Established Shapiro herself, who wrote How to Clone a Mammoth in 2015 and took a three-year sabbatical from UC Santa Cruz in March 2024 to become Colossal's Chief Scientific Officer, holds publicly that dinosaur de-extinction is impossible and that the real value of the technology lies in editing extant species for disease resistance, restoring lost genetic variation and correcting deleterious sequences. Frontier That is a defensible position and it is not the position the company's marketing occupies.

Frontier The thylacine is the scientifically serious programme and gets the least attention. A draft whole genome was assembled in 2017 from an ethanol-preserved pouch young at Museums Victoria; in 2023 RNA was recovered from a 130-year-old specimen in Sweden, the first RNA ever recovered from an extinct species and a genuine methodological advance, because RNA reveals tissue-specific expression that DNA cannot. Established A genome to 99.9 per cent completeness followed from a roughly 110-year-old skull in October 2024, with the complete genome about three months later. Frontier Colossal's partnership with the University of Melbourne dates from August 2022 and names the fat-tailed dunnart as closest living relative and intended surrogate — an animal weighing a few tens of grams standing in for a predator that weighed tens of kilograms. Frontier Alongside it the group reports mid-gestation marsupial embryo development in an artificial uterus. Speculative If full-term ex-utero marsupial gestation is achieved and independently replicated, it removes the surrogate and oocyte bottleneck for an entire clade, which would matter far beyond de-extinction; as of the record here it is one organisation's unreplicated result.

Frontier The most useful idea in the field barely gets called de-extinction, and it needs no technology at all. Two largely-coyote canids on Galveston Island, Texas carry red wolf alleles from a ghost population, and the proposal is to cross them into the captive red wolf population to restore lost variation. Established The wild red wolf population at Alligator River stood at fifteen to seventeen animals in 2023, rising to around thirty-five after releases that June, against roughly 270 in captivity. Speculative Generalised, this is the strongest exotic proposal in the subject: systematically survey admixed living populations for alleles from extinct taxa and breed them back into managed populations. Speculative It requires no editing, no cloning, no surrogate and no regulatory novelty; it addresses the actual conservation problem, which is loss of genetic diversity rather than loss of appearance; and nothing scientifically new has to be true for it to work. Frontier Its absence from the field's headline agenda is an attention failure, not a technical one.

Frontier The functional case dissolves the identity question, and it embarrasses the people making it. Pleistocene Park's measurement is that where animals trampled snow at minus 40°C air temperature, ground temperature was minus 30°C against minus 5°C under intact snow — a twenty-five degree difference in permafrost thermal regime produced purely by animal behaviour. Frontier A 2020 Oxford estimate suggested three experimental areas of a thousand animals each could sequester 72,000 tonnes of carbon over ten years. Established The park runs horses, reindeer, moose, muskox, yak, sheep, cattle, goats, camels and bison, and not one of them is a mammoth. Speculative If the mechanism is behavioural rather than taxonomic — and the trampling data suggest it may be — then the ecological argument for de-extinction argues for large herbivores generally, at a cost measured in livestock rather than in decades of reproductive biology. Handwave The version of the claim that survives contact with this data is not that we need a mammoth but that we need enough cattle, and no funder has been enthusiastic about that framing.

Frontier What is new since 2025 is that the scientific rebuttal now arrives with a DOI inside one publication cycle. Within months of the dire wolf announcement the subject acquired a peer-reviewed critical literature: Turner's argument in Ethics, Policy & Environment that the best case for de-extinction does not support this project, Hiller's welfare analysis in the same journal, a population-genetics verdict from Coltman and colleagues, and a set of commentaries by Rawlence, Hearn and Boer-Cueva and colleagues, each carrying a citable identifier. Established That makes the framing in which critics existed only on social media factually wrong. Frontier It sits uncomfortably beside a July 2025 report that academics critical of de-extinction research were targeted by an apparent coordinated campaign involving AI-generated blog posts and fraudulent copyright strikes, which the company denied involvement in. Speculative The two should be read as one story about how a capital-intensive science-adjacent enterprise absorbs criticism, not as two unrelated events.

Established The chain from an extinct genome to an animal that is genuinely the species has ten links, and the field has completed one of them. Link two, a complete reference genome with a catalogue of fixed differences against the living relative, is essentially finished: the thylacine genome is complete, the mammoth genome was done in 2015. Frontier That single completed link is why the field looks further along than it is. Established Link three is edit-scale escalation, from the current record of roughly twenty precise germline edits to the thousands or hundreds of thousands the genomic distance implies; this is an industrial-capability problem on a fast improving curve. Speculative Link four is germline mitochondrial genome replacement in a large mammal, which is not currently possible — a proxy built by editing a relative's nuclear genome carries the relative's mitochondria, permanently.

Established The binding link is link five, oocyte supply, and it is binding because reproduction has barely moved since 1996. Somatic cell nuclear transfer at scale needs oocytes; wild-species success remains under one per cent; there is no elephant or marsupial oocyte pipeline at anything like the required volume, and the only two routes out are induced-pluripotent-stem-cell-derived gametes or full ex-utero gestation, neither demonstrated in a large mammal. Frontier Link six, gestation, inherits the same problem plus a welfare cost that BioRescue has already paid once. Speculative Link seven, microbiome reconstitution, has no research programme at all: large-herbivore digestion is performed by a gut community rather than by the host genome, and coprolite DNA gives composition, not viable organisms. Handwave Link eight, behaviour and culture, may be unresolvable in principle for a socially learning species, because the proxy has no conspecific teachers and the knowledge was not written down anywhere except in the animals.

Frontier Two links are cheap and neither is being done. Link one is a falsifiable definition of success published before the animal exists — a stated percentage of species-diagnostic fixed differences, or a functional benchmark with a rate attached. Established No de-extinction programme has published such a specification, which means every outcome is a success by construction. Frontier Link nine, measuring the proxy's ecological effect against a matched control, is testable today with existing animals at Pleistocene Park scale for a small fraction of the capital already raised. Frontier Link ten, regulatory approval to release an engineered organism into a wild ecosystem, has no framework in any jurisdiction. Speculative That the field allocates to link three rather than to links one and nine is the most informative fact about what it is optimising for.

Established This subject waits on results that other fields are producing for their own reasons, and on one that nobody is producing at all. The genomic inputs are done or nearly so and are not the dependency: ancient-DNA extraction, contamination control, damage-pattern correction and assembly from short degraded fragments are mature methods, and what limits a paleogenome now is specimen quality rather than technique. Established DNA in a preserved specimen fragments progressively and deaminates at cytosines, which is why coverage rather than sequence chemistry sets the ceiling: two dire wolf individuals at 3.4× and 12.8× is what the specimens permitted, not what the instruments could do. Frontier The reproductive inputs come from assisted-reproduction research aimed at livestock and at human fertility: in-vitro-derived gametes from induced pluripotent stem cells, and the ex-utero gestation work assessed in Artificial Wombs, whose marsupial variant is the one route that would unblock the binding link without solving oocyte supply directly.

Frontier The multiplex editing capability comes from elsewhere entirely. Moving from tens of edits to thousands with verified on-target fidelity in a large-mammal germline is a problem owned by Genetic Engineering and by the synthesis and assembly capability discussed in Synthetic Biology; nothing about de-extinction accelerates it, and de-extinction will inherit it whenever it arrives. Established That is a real asymmetry in the field's incentives, because the capability it advertises is the one it is least responsible for developing. Speculative The dependency that is genuinely this subject's own is the one nobody has claimed: assembling a functioning gut community for an animal whose community is extinct, which is the community-assembly problem stated in Synthetic Ecosystems and the design problem stated in Microbiome Engineering. Frontier Neither literature currently cites the other, which is a solvable failure and a cheap one to fix.

Speculative The first required experiment is not an experiment and costs nothing. Publish, before any proxy animal exists, the quantitative criterion by which the claim would fail: a percentage of species-diagnostic fixed differences that must be present, or a named ecological function performed at a stated fraction of the extinct species' inferred rate. Frontier Without a pre-registered criterion, the dire wolf episode repeats for every subsequent species, because a phenotype claim and an identity claim can be swapped after the fact.

Frontier The genomics experiment that is actually missing is a fixed-difference catalogue at defensible coverage. Two paleogenomes at 3.4× and 12.8× cannot resolve most heterozygous sites; the specification is high-coverage, phased, chromosome-level assemblies from multiple individuals, with an explicit list of variants diagnostic of the extinct taxon against the chosen chassis. Frontier Against such a list, an edit count becomes a percentage, which is exactly why no programme has produced one. Speculative The same catalogue would settle the phylogenetic dispute the two dire wolf primaries have left open, since deep isolation and a two-thirds/one-third hybrid origin make different predictions about which variants are shared with which living canid.

Frontier Then the reproductive sequence, in order. Demonstrate a thousand or more verified precise germline edits in a large-mammal cell line with whole-genome off-target sequencing. Speculative Derive functional oocytes from induced pluripotent stem cells in a proboscidean or a marsupial and show fertilisation. Speculative Complete full-term ex-utero gestation in a marsupial and have it replicated by a second, unaffiliated laboratory — the single highest-value experiment in the subject, because marsupials have the shortest gestation and pouch-based development, and because success there converts a scientific unknown into an engineering programme.

Frontier Two experiments outside the laboratory would settle the arguments that the laboratory cannot. Run the Pleistocene Park design as a controlled trial: matched plots, ordinary livestock as the control arm, permafrost temperature and carbon flux as endpoints, and a pre-registered effect size. Speculative And run the moral-hazard study — measure stated willingness to pay for conservation, and legislative behaviour, before and after de-extinction publicity. Handwave That claim is made confidently on both sides, it is a cheap survey-and-natural-experiment design, and as far as this brief can establish nobody has run it.

Established The decisive infrastructure in the field's one success was a freezer. Willa's cells sat in the San Diego Zoo's Frozen Zoo for roughly thirty-five years before anyone could use them, and the result of that storage was a measurable increase in a living species' genetic diversity. Frontier Cryobanking is unglamorous, cheap relative to everything else in this brief, and the only intervention here that is unambiguously worth doing for species that are still alive. Speculative The engineering requirement that follows is a systematic obligation to bank viable cell lines from every species with a declining population, on the reasoning that a cell line is an option that costs a few dollars a year and expires only if the freezer fails.

Frontier Downstream of the freezer the engineering is reproductive, not genomic. The stack needs species-specific oocyte maturation protocols, nuclear transfer optimised for a species with no prior embryo-transfer literature, surrogate selection and welfare monitoring, and — for marsupials — a perfusion-and-support system able to carry an embryo to term outside a pouch. Established Each of those is a per-species engineering programme; almost none of it transfers between clades, which is why domestic-species cloning success rates of seven or eight in ten say nothing about elephants. Frontier Editing hardware is the least constrained part of the chain: the capability to make twenty verified changes exists, the capability to verify a thousand is a sequencing and analysis problem rather than a biological one, and the current record is limited by what anyone has needed rather than by what can be built.

Speculative The one piece of hardware that would change the shape of the problem is an artificial marsupial pouch. A perfusion system carrying a fat-tailed dunnart-sized embryo from mid-gestation to independence would replace a surrogate whose body mass is three orders of magnitude below the target animal's, and would do so in the clade where gestation is measured in weeks rather than years. Handwave The elephant version of the same device — a 22-month, several-hundred-kilogram gestation supported entirely ex utero — is not an extension of the marsupial one; it is a different engineering discipline that does not yet exist and has no funded precursor.

Established The technology cluster around de-extinction is more valuable than de-extinction. Ancient-DNA methods developed for extinct taxa now underwrite conservation genomics for living ones, including the detection of ghost alleles in admixed populations; the 2023 recovery of RNA from a thylacine specimen extends that toolkit to expression, which no amount of DNA can supply. Frontier Editing for disease resistance in extant threatened species is the same technology pointed at a tractable target: engineering cane-toad-toxin resistance into the northern quoll, reported in May 2024, is the most defensible item in Colossal's portfolio precisely because it is not resurrection.

Frontier Three adjacent programmes would each be unblocked by a result this field needs anyway. Ex-utero gestation, pursued here for marsupials, is the same capability assessed in Artificial Wombs and would transform captive breeding for any species with a difficult pregnancy. Frontier Induced-pluripotent-stem-cell gametogenesis would end oocyte scarcity for every endangered-species breeding programme, not only for proxies. Speculative Synthetic gut-community assembly, needed here to feed a hypothetical proxy on a diet its chassis did not evolve for, is the same problem as building a closed loop in Synthetic Ecosystems. Frontier In each case the de-extinction framing is the fundraising wrapper and the adjacent capability is the payload.

Frontier Cryobanking is the adjacent technology with the best ratio of value to cost and the least attention. The Frozen Zoo's holdings turned into a living, reproducing ferret line thirty-five years after the cells went in, which is the only demonstrated case anywhere of biobanking paying off at that timescale in a vertebrate. Speculative Extending the same practice to genome-resolved tissue archives — cell lines rather than DNA alone, because only cell lines support nuclear transfer — is a logistics and funding problem with no scientific unknowns in it. Handwave At the far end sits the proposal that a sufficiently complete tissue archive plus mature reproductive technology makes the current extinction pulse partially reversible by a later century, which is coherent, entirely undemonstrated, and the strongest reason to keep the freezers running whatever one concludes about proxies.

Established There is a standard, it was written in advance, and it was applied. The IUCN Species Survival Commission created a De-extinction Task Force in April 2014 and published guiding principles on creating proxies of extinct species for conservation benefit in 2016; the Canid Specialist Group measured the dire wolf against them and found the animals are not a dire wolf under any definition, that the project does not contribute to conservation, and that phenotypic proxies do not alter an extinct species' status. Frontier That is what a functioning standard looks like — it produced a verdict within weeks of an announcement — and it has no enforcement mechanism whatsoever.

Established The regulatory gap is at the other end and is total. No jurisdiction has a framework for approving the release of a deliberately engineered vertebrate into a wild ecosystem. Frontier The nearest analogues — engineered-organism release rules written for insects and plants, and endangered-species listing law written on the assumption that extinction is terminal — both break on a proxy: an animal that is legally not the extinct species cannot inherit its protections, and an animal that is legally the extinct species would trigger critical-habitat obligations nobody has budgeted for. Speculative The question of whether a listed species can be delisted because a proxy exists is currently unanswerable in law, and it is the question that makes the moral-hazard argument more than rhetoric.

Frontier The funding asymmetry is the institutional fact that explains the rest. A private company raised $435 million at a $10.2 billion valuation for resurrection, while the programme that measurably increased a real endangered species' genetic diversity ran on zoo, non-profit and federal agency budgets. Speculative On the sympathetic reading, the resurrection framing is a wrapper that pulls capital into a chronically starved sector and pays for a bioinformatics spin-out, marsupial artificial-uterus work, quoll toxin resistance and red wolf cloning that no conservation grant would have funded. Speculative On the unsympathetic reading, the wrapper's cost is public misunderstanding of whether extinction is reversible, and that cost is borne by every conservation programme rather than by the company. Frontier Both readings are consistent with the evidence and this brief declines to pick one.

Established The welfare cost is not hypothetical and has already been paid. Cloning at sub-one-per-cent efficiency in wild species means most attempts end in failed pregnancies, and BioRescue lost a surrogate, Curra, to a bacterial infection after a successful embryo transfer in 2023. Frontier The welfare argument, made in the peer-reviewed literature by Hiller in 2025, is the part of the critique that the conservation objection tends to skip: the animals produced are experimental subjects with no conspecifics, no species-typical social environment and, in the mammoth case, a proposed life in a climate their chassis did not evolve for.

Frontier The deception argument is separate from the welfare one and is stronger than it looks. Turner's objection is that the best philosophical case for de-extinction — restitution for a wrong humans committed — does not license a project that produces an animal which is not the wronged species. Speculative On that reading, an edited grey wolf presented as a restored dire wolf is not a partial success but a category substitution, and the harm is to the public's model of what extinction means rather than to any animal. Frontier The IUCN's stated worry that the effort may threaten extant grey wolves is the same argument in policy form: if extinction reads as reversible, the political case for prevention weakens.

Speculative The counter-argument deserves stating at full strength. Nothing about publicity for de-extinction has been shown to reduce conservation funding or weaken conservation law, and the claim that it does is an empirical proposition that its proponents have not tested. Frontier Meanwhile the black-footed ferret case shows that cloning from cryobanked cells produces real conservation benefit, and the quoll work shows the same for editing. Handwave If the moral-hazard effect is small and the capital effect is large, the honest verdict flips — and the reason nobody knows which way it flips is that the study costs less than one month of the programme it would assess.

Established The civilizational question is whether extinction is a boundary or a cost. Every conservation institution built since the nineteenth century assumes the first: that a species lost is lost permanently, which is what makes prevention worth paying for. Frontier De-extinction proposes, without yet demonstrating, that it is the second — expensive, difficult, but in principle reversible. Speculative On the evidence assembled here the proposition is false as stated: no programme has restored a species, one restored a subspecies for minutes, and the credible near-term products are proxies that resemble their targets from the outside.

Speculative What a genuine restoration would require is worth stating plainly, because it clarifies the scale. The full nuclear genome edited or synthesised rather than sampled, on the order of tens of millions of changes; mitochondria replaced; a heritable gut community reconstituted from organisms that no longer exist; correct epigenetic state through a process notorious for getting imprinting wrong; and behaviour reconstituted in an animal with no teachers. Handwave The last of those may be unresolvable in principle for a culturally transmitting species, which means that for elephants, whales and corvids there is a version of extinction that no technology reverses, only re-approximates.

Frontier The durable civilizational asset in this subject is the freezer, not the animal. Cryobanked cell lines from species alive today are cheap, they preserve options that cannot be recovered later, and they have already produced the field's only unambiguous win. Speculative A civilization that banks tissue systematically for a century, and never resurrects anything, will have done more for biodiversity than one that produces a proxy every decade. Frontier The asymmetry is that banking is reversible in the useful direction — an unused cell line costs a freezer slot, an unbanked species costs everything — and that the decision has to be made while the animal is alive, which is the one window this field's headline programmes do not address at all.

These horizons track proxies, capabilities and the one thing that would change the argument — a reproductive result, not a genomic one.

• 10 yr: Frontier More proxies, on the dire wolf pattern: tens of edits in a living relative, announced as a species. Expect a mammoth-phenotype elephant embryo before a calf, and expect the calf to slip past the stated 2028 date. Frontier Cloning from cryobanked cells extends to several more endangered species on the black-footed ferret template, which is the line of work most likely to produce a genuine conservation result in this window. Speculative Full-term ex-utero marsupial gestation is demonstrated or abandoned; either outcome is decisive.
• 25 yr: Speculative If ex-utero gestation and induced-pluripotent-stem-cell-derived oocytes both land, the reproductive wall falls and edit count becomes the only constraint, which is the good case. Speculative A thylacine proxy is the most plausible first product of that route because the surrogate problem is soluble in marsupials. Handwave A mammoth proxy released onto tundra remains unlikely in this window on gut-community grounds alone, since nobody is working on the digestive half of the problem. Speculative Expect the first serious regulatory fight over releasing an engineered vertebrate into a wild ecosystem in this window rather than the last one, because no jurisdiction has a framework and the first application will have to create it.
• 50 yr: Speculative Genome writing at the scale the identity claim requires — millions of changes rather than tens — becomes technically conceivable, at which point the species-concept argument becomes empirical rather than rhetorical for the first time. Handwave Behaviour and culture remain unreconstructed, so the products are anatomically convincing animals with no inherited knowledge, which is a new kind of entity rather than a restored old one. Handwave Mitochondrial replacement at germline scale is the quiet condition on all of this and it has no research programme today, which means the fifty-year proxy may still carry its chassis species' mitochondria and its chassis species' maternal line.
• 100 / 250+ yr: Handwave On the optimistic branch, banked tissue from the current extinction pulse becomes the raw material for large-scale restoration by descendants with better reproductive biology, and this century's cryobanks turn out to have been its most consequential conservation investment. Handwave On the pessimistic branch the freezers fail, the funding lapses, and the whole enterprise is remembered as a period in which humanity told itself extinction was reversible while it accelerated.

• Depends on This brief waits on Synthetic Ecosystems for a result nothing else on this map produces: the ability to assemble a functioning microbial community to specification and establish it in a host. A large herbivore's digestion is performed by a gut community rather than by its genome, and a mammoth proxy would inherit an Asian elephant's community, adapted to tropical browse, and be released onto Arctic graminoids. Coprolite DNA yields composition, not viable organisms, so there is no ancient community to recover and one would have to be built. That is precisely the closed-loop community-assembly problem, and it is the only link in this subject's chain with no research programme attached to it at all.
• Enables What this subject supplies to others is a set of hard numbers that any resurrection or restoration claim now has to clear, and a naming convention that makes such claims falsifiable. The reproductive record — sub-one-per-cent somatic cell nuclear transfer success in wild species, 277 eggs for one adult sheep, 814 attempts for one horse, 38 cryopreserved rhinoceros embryos and no birth — is the reference series against which any proposal to reconstitute an animal from stored material should be assessed. The black-footed ferret result supplies the positive benchmark: cells frozen for thirty-five years, one cloned animal, three times the genetic diversity of the living population, and confirmed reproduction in the clone line. It also exports a naming discipline that any adjacent field making identity claims about engineered organisms could adopt at no cost: the Quagga Project's decision to call its animals Rau quaggas converts an unfalsifiable claim about what an animal is into a falsifiable claim about what it looks like, and every dispute recorded in this brief would have been shorter if that convention had been general. No typed enabling edge is recorded, because what this subject exports is a standard of evidence rather than a component.
• Adjacent The adjacent subjects share components and, more usefully, share bottlenecks. Artificial wombs own the ex-utero gestation capability that would unblock this subject's binding link, and marsupials are the clade where that capability is most likely to be demonstrated first. Genetic engineering owns the multiplex editing scale that would let an edit count become a percentage of a species-diagnostic difference set. Microbiome engineering owns the community-design methods that the digestive problem needs. Synthetic biology owns whole-genome synthesis, which is the only route to the version of this subject in which the identity claim is true rather than approximate. In every case the capability is being developed for other reasons and this subject is a consumer of it.

Established “Colossal brought back the dire wolf.” Twenty edits across fourteen genes were made in grey wolf cells and no ancient dire wolf DNA was inserted. Established The company's own Chief Scientific Officer calls the animals grey wolves with twenty edits; the IUCN Canid Specialist Group calls them not a dire wolf under any definition. Established Against a genomic distance between the taxa measured in millions of sites, twenty edits is not an approximation of a species — it is a phenotype selection, and a good one.

Established “There was a scientific paper behind the announcement, so the claim was peer-backed.” There was a paper and it is worth reading precisely because it does not make the claim. Frontier Gedman and colleagues is a phylogenomics preprint reporting two paleogenomes at 3.4× and 12.8× coverage and eighty genes under diversifying selection, placing the dire wolf outside Canis and proposing a hybrid origin. Established It is not peer reviewed, it contains no de-extinction claim, and its candidate gene set is nearly six times larger than the set that was edited. Established The peer-reviewed prior, Perri and colleagues in Nature, puts the split from living canids at about 5.7 million years ago with no evidence of gene flow into grey wolves or coyotes. Frontier The genomics does not support the premise of the engineering, and both primaries say so in different ways.

Established “Dire wolves and grey wolves are closely related, so twenty edits is a reasonable start.” The number to print is 5.7 million years. Frontier The two primaries disagree about whether there was ancient hybridisation and agree that the dire wolf lineage sits outside the genus containing the grey wolf, which is why the choice of a grey wolf chassis is a practical decision about available cell lines rather than a claim about proximity.

Established “The woolly mouse proved mammoth cold adaptation.” Cold tolerance was never tested — an objection made by Robin Lovell-Badge and confirmed by the preprint, whose reported phenotypes are curly, textured coats and golden-brown hair and nothing else. Established The preprint's own figure is simultaneous editing of up to seven different genes, not the twelve usually quoted; the larger count appears to aggregate a gene set across lines rather than to describe any one animal. Frontier What the paper actually demonstrates — a rapid multiplex platform for testing mammoth-derived variants — is a real and useful result, and it is a smaller one than the headline.

Established “De-extinction has never worked.” It has, once. Established The Pyrenean ibex was cloned in 2003 from cells of Celia, the last individual, who died in 2000; the clone died shortly after birth of a lung defect, and a second attempt began in 2013 to check whether the frozen cells remained viable. Established The honest statement is that de-extinction has succeeded exactly once, briefly, in a subspecies with cryopreserved cells taken from a living animal — which is the easiest case that will ever exist.

Established “Cloning is solved, so the hard part is the genome.” This is exactly backwards, and it is the enthusiast's error rather than the sceptic's. Established Dolly took 277 eggs for 29 embryos for three births for one survivor; Prometea took 814 attempts; success in wild species remains under one per cent. Established The thylacine genome, meanwhile, is complete. Frontier Genomics is the solved part of this stack and reproduction is the wall, and because the genome is essentially solved, enthusiasts infer that the rest will follow — when mitochondria, oocytes, gestation, microbiome and behaviour are each independently unsolved and several have no active research programme at all.

Established “Back-breeding recreates the extinct animal.” The aurochs literature is explicit that breeding for an aurochs-like phenotype does not equate to an aurochs-like genotype, and back-bred animals are phenotypic look-alikes often lacking the behavioural, physiological and ecological traits of the original. Established The Quagga Project encodes the correct position in its own vocabulary by calling its animals Rau quaggas and stating that they differ genetically from the quagga.

Established “Heck cattle were Nazi pseudoscience with no scientific content.” Half right. Established The programme was condemned by the Union of German Zoo Directors in 1951 and is inseparable from its ideological context. Established But its core genetic premise — that aurochs alleles persist in extant cattle — was later confirmed by sequencing a 6,750-year-old aurochs bone from England, which showed variants shared with British and Irish breeds. Frontier The premise was sound and the claim was overstated, which is the same failure mode as the dire wolf, eighty years earlier.

Established “Colossal is a de-extinction company.” By output it is closer to a genome-editing and reproductive-technology company. Established It spun out Form Bio with $30 million, engineered cane-toad-toxin resistance into an extant threatened marsupial, is cloning red wolves, established an Avian Genomics Group in January 2023 pairing dodo sequence with Nicobar pigeon primordial germ cells, and reports mid-gestation marsupial development in an artificial uterus. Frontier The extant-species work is the most defensible part of the portfolio and receives the least coverage, which is a marketing decision rather than a scientific one.

Established “A mammoth by 2028.” Set it against the benchmark. Established BioRescue has 38 cryopreserved northern white rhino embryos, two living females of the subspecies, a closely related surrogate species with established embryo-transfer protocols — and no live birth, plus a surrogate lost to infection in 2023. Frontier Elephant reproductive technology is considerably less developed than rhinoceros reproductive technology, gestation is around 22 months, and no elephant oocyte pipeline exists. Speculative A calf by 2028 would require the reproductive biology of the last thirty years to be wrong, in a direction nobody has reported.

Handwave “A proxy that does the ecological job is good enough, so the identity argument is beside the point.” This is the strongest position available to the field and it has an awkward consequence its advocates rarely state. Frontier Pleistocene Park's own data attribute the permafrost effect to trampling behaviour — ground at minus 30°C against minus 5°C under intact snow — and the park is testing the thesis with horses, reindeer, muskox, yak, cattle and bison. Speculative If function is behavioural rather than taxonomic, the functional argument for de-extinction is an argument for stocking large herbivores now, not for building a mammoth by 2038. Handwave The proxy defence, taken seriously, dissolves the case for the technology it was invented to defend.