Forecast
Heritable human genetic enhancement beyond disease correction occurs in at least one jurisdiction by 2126.
Last revised Sep 11, 2026.
Current estimate
probability this has happened by each date
A verified live birth of a human carrying an intentional heritable edit at a locus where that embryo carried no disease-associated variant, and where the modification is not a recognised treatment or prevention of a condition the embryo was at elevated risk of. Assessment is from the genetic and clinical facts, not from the operator's stated purpose.
Resolution criteria
- 2126
- A verified live birth of a human carrying an intentional heritable edit at a locus where that embryo carried no disease-associated variant, and where the modification is not a recognised treatment or prevention of a condition the embryo was at elevated risk of. Assessment is from the genetic and clinical facts, not from the operator's stated purpose.
Probability history
| Revised | State | p(2035) | p(2050) | p(2100) |
|---|---|---|---|---|
hf-forecast-0024 | researched | — | — | — |
hf-forecast-0007 | initial prior | — | — | — |
Strongest evidence for
- It has already happened. Three children were born with intentional heritable edits, confirmed by official investigation, within about six years of CRISPR becoming usable, carried out by one funded individual against universal condemnation (hf-evidence-0042).
- Deterrence was real but not permanent: the actor served three years, was fined three million yuan, and has since resumed research and sought a permissive jurisdiction (hf-evidence-0042).
- Ten of 106 surveyed countries have no relevant policy at all, and the forecast needs one (hf-evidence-0043).
- Two arguments that appear to oppose the forecast do not survive inspection. Polygenic selection's ceiling of about 2.5 IQ points or 2.5 centimetres is too low to absorb demand for large enhancement, and the most-cited safety statistic against CCR5 editing was retracted in 2019 over a genotyping artifact (hf-evidence-0046).
- Commercial appetite for engineering non-disease traits in offspring is demonstrably real today, sold as polygenic selection.
Strongest evidence against
- No country explicitly permits heritable editing and 75 of 96 with any policy prohibit initiating a pregnancy with a modified embryo (hf-evidence-0043).
- The traits people most want cannot be edited: height involves 12,111 significant SNPs across 21 percent of the genome, explaining only about 40 percent of variance (hf-evidence-0044).
- Embryo editing still produces mosaicism in roughly 14 to 25 percent of embryos, structural damage in about 16 percent of cells, and widespread loss of heterozygosity (hf-evidence-0045).
- Legal incentive pushes every actor to frame work as therapy, which suppresses the clean enhancement case even where enhancement occurs.
Present uncertainties
- Whether a clean enhancement case ever becomes distinguishable from a therapy-framed one, and whether multiplex editing ever makes polygenic traits addressable.
Next discriminating observation
A heritable edit at a locus where the embryo carried no disease-associated variant, at any stage from approval to birth; and any jurisdiction moving from silence to explicit permission.
Sources
- The one time heritable human editing actually happened
Three children were born carrying intentional heritable CCR5 edits, confirmed through the Chinese official investigation rather than through any publication. He Jiankui was sentenced to three years in prison and fined three million yuan for illegally carrying out human embryo gene editing for reproduction. His manuscript was submitted to and rejected by both Nature and JAMA, so the molecular claims were never peer reviewed. Since release he has resumed research activity and is reported to be seeking a permissive jurisdiction for a further attempt.
Limit: The stated purpose was HIV resistance, which is disease prevention rather than enhancement, so this precedent does not itself resolve the forecast. The molecular details rest substantially on a leaked, never-peer-reviewed manuscript.
Retrieved by a delegated source worker and not independently re-fetched.
- Human germline and heritable genome editing, the global policy landscape
Of 106 countries surveyed, 96 have a relevant policy document. Seventy-five of those 96 prohibit using genetically modified in vitro embryos to initiate a pregnancy. No country explicitly permits heritable human genome editing. Ten of the 106 have no relevant policy at all. For laboratory germline research specifically, 23 prohibit and 11 explicitly permit.
Limit: Absence of a policy is not affirmative permission, and the survey does not name the ten countries. A 2020 snapshot of a landscape that moved sharply after 2018 and will move again across a century.
Fetched and confirmed directly by the master agent.
- A saturated map of common genetic variants associated with human height
Across 5.4 million individuals, 12,111 near-independent SNPs reach genome-wide significance for height, spread across 7,209 genomic segments covering roughly 21 percent of the genome, jointly explaining about 40 percent of height variance in European-ancestry prediction and only 10 to 20 percent in other ancestries. No single locus or small set of loci accounts for a meaningful fraction of the trait.
Limit: Common-variant architecture only; it does not exclude rare large-effect variants, and it says nothing about multiplexed editing of many loci at once across a century. The forecast needs one trait, not height.
Retrieved by a delegated source worker and not independently re-fetched.
- Unintended outcomes in human embryo genome editing
Published human embryo editing studies report mosaicism in roughly 14 to 25 percent of edited embryos, unintended structural damage in about 16 percent of cells, and loss of heterozygosity in a majority of deeply sequenced samples in one study. Targeting efficiency in the best-reported case was 72.2 percent.
Limit: Measured on 2015 to 2021 double-strand-break technology. Base and prime editing were developed specifically to address these failure modes, so these figures describe a moving target rather than a fixed limit.
Retrieved by a delegated source worker and not independently re-fetched.
- Polygenic selection has limited utility, and the flagship editing-risk statistic was retracted
Selecting the highest-scoring embryo from a typical IVF batch yields a mean expected gain of about 2.5 centimetres of height or about 2.5 IQ points at current polygenic score performance, and in real families the child with the highest height score was the tallest sibling only 25 percent of the time. Separately, the most-cited risk statistic against CCR5 editing, a 21 percent increase in mortality for homozygotes, was retracted by Nature Medicine on 8 October 2019 after a genotyping calling bias in UK Biobank data was found to explain the result.
Limit: Selection gains rise as polygenic scores improve, so the ceiling is not fixed across a century. The retraction removes one specific claim of harm; it does not establish that CCR5 editing is safe.
Fetched and confirmed directly by the master agent.