For a long time, the idea of being able to choose certain genetic characteristics of a future child belonged to the realm of science fiction.
That is no longer entirely the case.
In the United States, new companies are now offering couples undergoing in vitro fertilization (IVF) the possibility of analyzing their embryos' DNA to estimate their genetic risk of developing certain diseases. Some go even further, offering estimates related to height, physical characteristics and even complex traits such as cognitive abilities.
We are not yet in the age of “designer babies.” But for the first time, the line between understanding our genetics, selecting based on them, and potentially modifying them is becoming far less theoretical.
And artificial intelligence could significantly accelerate this evolution.
Genetic testing of embryos is not new.
For several years, it has been possible, as part of IVF, to screen for certain chromosomal abnormalities or mutations responsible for serious single-gene disorders.
In these situations, the relationship between a genetic variant and a disease can be relatively direct.
What is changing today is the use of polygenic risk scores.
Conditions such as type 2 diabetes, cardiovascular disease and many cancers generally do not depend on a single gene. They are influenced by hundreds, or even thousands, of genetic variants, along with environmental factors, lifestyle and many other variables.
Algorithms can analyze these variants simultaneously and produce a statistical risk score.
When applied to several embryos from the same IVF cycle, the principle becomes very different: instead of simply detecting a specific mutation, it becomes possible to compare the genetic probabilities associated with each embryo.
One embryo might have a slightly lower estimated genetic risk of cardiovascular disease. Another might have a statistically more favorable profile for another condition.
And this is precisely where the debate begins.
Our ability to interpret the human genome remains imperfect. But it is progressing rapidly.
Genetic databases are growing, studies involving millions of individuals are identifying more associations between genetic variants and human characteristics, and artificial intelligence is making it possible to analyze extremely complex relationships within these data.
AI does not change our DNA.
What it changes is our ability to interpret it.
And as that interpretation becomes more precise, the amount of information potentially available about an embryo increases.
Today, the discussion focuses primarily on the risk of certain diseases. Tomorrow, it could extend to a much broader range of biological characteristics.
It is this progression that transforms a medical question into a genuine societal one.
This is where an important scientific distinction is necessary.
A genetic score is not a certain prediction.
Even for a trait strongly influenced by genetics, such as height, genes represent only part of the equation. For much more complex characteristics such as intelligence, personality or abilities, the interactions between genetics, development, education and environment are considerable.
Parents are also limited by the embryos they have actually produced. This is not about choosing any imaginable genetic combination from the human population, but rather comparing a small number of embryos produced by the same two parents.
Another important limitation is that today's genetic datasets represent some populations much better than others, particularly people of European ancestry. The accuracy of these scores can therefore vary considerably depending on a person's genetic ancestry.
Finally, statistically reducing a risk does not mean eliminating it.
A decrease in risk from 4% to 3%, for example, represents a 25% relative reduction, but only one percentage point in absolute risk.
How these results are communicated to prospective parents may therefore be almost as important as the technology itself.
Despite the arrival of companies already offering these services, several major medical organizations remain highly cautious.
In 2026, the American Society for Reproductive Medicine still considers polygenic embryo screening an emerging technology whose clinical utility has not been sufficiently demonstrated for routine use.
Among the concerns raised are the still-limited accuracy of predictions, the lack of diversity in genetic databases, our incomplete understanding of interactions between genes and the environment, and the possibility that selecting to reduce one genetic risk could affect others.
In other words, the ability to generate a score does not necessarily mean we yet know how to use it appropriately.
Another distinction is essential.
The technologies currently being discussed do not modify an embryo's genes.
They allow several embryos to be compared before deciding which one will be transferred.
Genetic editing represents an entirely different step.
With technologies such as CRISPR, it is theoretically possible to directly modify a DNA sequence. A modification made in an embryo could then be present in virtually every cell of the child and, if it affects the germline, potentially be passed on to future generations.
We would then move from:
“Which embryo should we choose?”
to:
“What are we prepared to modify?”
And that second question is far more profound.
Imagine that one day a technology could safely correct a mutation that would otherwise cause a child to develop a serious genetic disease.
Many people would probably consider its use desirable.
But what comes next?
If we can reduce the risk of Alzheimer's disease, why not depression?
And what if we could influence height?
Muscle mass?
Memory?
Certain cognitive abilities?
The line between preventing disease and enhancing a human being could become extremely difficult to define.
There is also an economic question.
Advanced reproductive technologies are currently expensive. If certain interventions eventually make it possible to significantly reduce disease risks or influence certain characteristics, they will probably first be accessible to wealthier families.
Inequality could then extend beyond education, wealth or access to healthcare.
It could begin before birth.
On the other hand, as we have already seen with genetic sequencing, the cost of these technologies could fall rapidly as they become more widely adopted.
What seems accessible to only a handful of families today could become far more common within one or two decades.
Perhaps the most interesting question is not whether we will one day have “designer babies.”
That dramatic expression oversimplifies the scientific reality.
The real transformation is more subtle.
Over the past few decades, we have learned to read DNA.
We are now learning to better interpret it, particularly with the help of artificial intelligence.
We are beginning to use that information to select.
And science is already exploring ways to modify.
Read. Understand. Select. Modify.
Each step brings extraordinary possibilities — and questions that technology alone cannot answer.
As our ability to understand the genome advances, the question may soon no longer be simply:
“What can we learn from our DNA?”
But rather:
“What do we want to have the right to do with it?”
Other Journal: News
VIEW MORE
