STUDIO TORIUMI / OBSERVATORY
In 2018, gene-edited babies were born in China. In the United States, people with no laboratory affiliation have injected CRISPR into their own arms. And in animal facilities, mice have been born from two same-sex parents. All three are real. But in every case the reported version and the actual one differ substantially. This page records what happened, and what didn't — including the parts that are inconvenient.
Gene-editing coverage routinely blurs these two. Separate them and the rest of the field becomes legible.
A somatic edit is carried by the person who chose it, even when it fails. Personal responsibility actually applies.
Germline is different. The person edited does not yet exist and cannot refuse — and the change then passes to their children and grandchildren. There is nobody available to ask. That, rather than the technical risk, is why this is called a line.
What the "gene-enhanced babies" story actually was. It was not an enhancement attempt. That is the central misconception.
He Jiankui, then at the Southern University of Science and Technology, announced days before the Hong Kong genome-editing summit that twin girls had been born from edited embryos. A third child followed in 2019.
The target was CCR5, the gene encoding a protein HIV uses to enter human cells. Disable it, the reasoning went, and the child resists HIV. The couples recruited had HIV-positive fathers.
In December 2019 he was sentenced to three years in prison and fined three million yuan for illegal medical practice. He was released in 2022.
1. There was no medical need. Sperm washing already prevents transmission from an HIV-positive father. Nothing justified the risk.
2. The intended variant was not produced. Natural HIV resistance comes from CCR5-Δ32, a specific deletion. What was actually created were different, novel variants whose function nobody knows.
3. Mosaicism remained. Not every cell was edited — the children carry a mixture of edited and unedited cells.
4. Losing CCR5 has costs. It is associated with worse outcomes in West Nile virus and influenza. One risk was traded for another.
A 2016 UCLA mouse study suggested that removing CCR5 improved memory, which fuelled reporting that He was pursuing cognitive enhancement. He denied it, and no evidence supports it. But that paper is why the "designer baby" framing stuck.
The reverse claim collapsed too. In 2019 a Nature Medicine paper reported that people homozygous for CCR5-Δ32 had 21% higher mortality. It was retracted the same year — a genotyping error in UK Biobank data.
So the case for benefit and the case for harm both lost their footing. An irreversible edit was made on top of things nobody actually knew. That is the substance of this episode.
This is the somatic side — it stops with them. No consent problem. A different problem instead: almost nothing happened.
A former NASA researcher and founder of The ODIN, which sells DIY biology kits. At a conference he injected himself with CRISPR designed to knock out myostatin, the gene that limits muscle growth. In theory, more muscle.
Nothing happened. The edit reached a small number of cells near the needle and produced no measurable change. Zayner has since publicly said he regrets encouraging others to copy it.
An adult body has tens of trillions of cells. An injection edits the ones near the needle. Editing a single fertilised egg and editing a whole adult differ by many orders of magnitude. That one fact explains most of why biohacking looks dramatic and produces nothing.
An HIV-positive man injected himself with an unapproved gene therapy while streaming it, intending to make his body produce a neutralising antibody.
His viral load did not fall. He later returned to standard antiretroviral treatment.
As CEO of Ascendance Biomedical, he injected an untested herpes therapy into his own thigh on stage at a Texas conference. His company, which sought to supply unapproved treatments to terminally ill people, was already under heavy criticism.
In April 2018 he was found dead at 28, in a flotation tank. The death was ruled accidental drowning. No link to the injection was established — but the episode is remembered as the high-water mark of the scene's disorder.
The CEO of BioViva travelled to Colombia, where approval was not required, and received telomerase and follistatin gene therapy. She reports that her telomeres lengthened.
The work was never peer-reviewed, had no control group, and the measurement method has been challenged. As it also promotes a company she runs, specialists generally treat it as unusable as evidence.
In November 2017 the FDA stated plainly that selling gene therapy kits for self-administration is illegal. California passed a law in 2019 requiring DIY CRISPR kits to carry a notice that they are not for self-administration.
This isn't a lecture, it's a finding: among publicly documented self-experiments, none has demonstrated the intended effect. High risk, no return.
Mice, not humans. But of everything on this page, this may be the work with the longest reach — it moves a precondition of reproduction itself.
29 mice were born carrying genetic material from two females only (bimaternal). They were healthy, and some went on to have offspring of their own.
The same study produced bipaternal pups from two fathers — but all died within 48 hours. Two mothers and two fathers are not equally difficult problems.
Mammals have genomic imprinting: certain genes are marked so that
only the paternal copy, or only the maternal copy, is expressed. Without both sets of marks, development breaks down.
What this study did was delete the imprinted regions responsible — three of them for the bimaternal case, seven for bipaternal. The father side needs more intervention, and the survival gap reflects that.
A different route entirely. Rather than deleting imprints, the team took male (XY) cells, converted them to XX in iPS culture, and grew functional eggs from them. Eggs, from a male.
Fertilised with sperm from another male, these produced pups descended from two fathers.
Efficiency was very low: seven pups from 630 embryos, roughly 1%. Hayashi has repeatedly said human application is a decade away at minimum, with no clear path on safety.
The bipaternal mice that died within 48 hours in 2018 were revisited by extending the intervention to around twenty imprinted regions, and some individuals reached adulthood.
Survival remains low and many animals show developmental abnormalities. The distance between "possible" and "safely possible" is still considerable.
It gets discussed in the context of same-sex parenthood, but what the researchers are actually attacking is a basic-biology question: why do mammals require both a father and a mother at all? They are testing the imprinting mechanism from the other direction.
Human application is far away, technically and legally. What changed after 2018 is only that it stopped being impossible in principle.
So far this page has been failures and recklessness. But the same technology, on the near side of the line, is already treating people. Leaving that out would be dishonest.
For sickle cell disease and beta thalassemia — conditions involving severe pain crises and transfusion dependence — the UK and US formally approved a CRISPR-based treatment.
The method: take the patient's own blood stem cells, edit them outside the body, and return them. The edit switches a fetal haemoglobin gene back on. Only the patient's own cells are changed; nothing is passed to children.
In trials, many patients had their pain crises disappear entirely. The obstacle now is price — on the order of two million dollars per patient.
An infant diagnosed shortly after birth with CPS1 deficiency, an extremely rare metabolic disorder, received a base-editing therapy designed, manufactured and administered for his specific mutation within months. Children's Hospital of Philadelphia.
What is new here is speed and specificity rather than technique. Diseases with a handful of patients have never had an economic case for drug development. Building one for a single person, fast enough to matter, is now demonstrably possible.
It sits at the exact opposite pole from He Jiankui: medically necessary, consented, somatic, inside the regulatory system, and published.
Details vary by country; the broad shape does not.
"Designer babies are already possible; ethics is what's holding it back." That is not accurate.
Traits like height and intelligence involve thousands of genes each contributing a minute amount. Rewriting one or two produces no measurable effect. Single-gene traits like the CCR5 case are the exception — and even in that one instance, the intended edit was not achieved.
It isn't only ethics holding the line. The capability isn't there. Which is precisely why the real present-day risk looks the way it does: irreversible edits being made on top of effects nobody can predict.
Only the words that trip people up.
Collecting these cases is neither endorsement nor condemnation. This field attracts both excessive hope and excessive fear, so the only goal here is to keep what was achieved, what wasn't, and what remains unknown in separate boxes.
Not medical advice. Self-administration is not recommended — setting ethics aside, no publicly documented self-experiment has produced its intended effect.
Principal sources — He Jiankui: Second International Summit on Human Genome Editing (Hong Kong, 2018) and Xinhua reporting on the verdict (Dec 2019) / The Nature Medicine CCR5-Δ32 mortality paper was retracted in 2019 / Li Z-K et al., Cell Stem Cell, 2018 (bimaternal and bipaternal mice) / Murakami K, Hamazaki N, Hayashi K et al., Nature, 2023 (oocytes from male cells) / Casgevy: MHRA Nov 2023, FDA Dec 2023 / FDA, Information About Self-Administration of Gene Therapy, Nov 2017