Can AI Design a Virus — and Could AI Save Us From It?
On 24 September, Dutch science journalist Maarten Keulemans asked a rather uncomfortable question in de Volkskrant: How realistic is it that AI designs a deadly virus and destroys humanity?
As questions go, it is not exactly something to read over breakfast.
A few years ago, I would have filed it comfortably under science fiction. Today, the answer is more complicated, because AI has already been used to design viruses. Not viruses that infect humans, fortunately, and certainly not a pandemic weapon. Researchers used generative AI to design bacteriophages, viruses that infect bacteria, and some of those computer-generated designs became functioning viruses when their DNA was synthesized and tested in the laboratory.
So perhaps Keulemans’s enormous question needs breaking into smaller ones.
Can AI design a virus?
Yes.
Can it currently design something capable of wiping out humanity?
There is no public evidence that it can.
The interesting territory lies between those two answers.
The same technology can defend us
Whenever we discuss engineered viruses, the conversation quickly arrives at bioterrorism.
That possibility cannot simply be dismissed. The World Health Organization has specific frameworks for research that could be misused, and AI companies are increasingly concerned about models assisting users with sophisticated biological work.
But there is an equally important side of this story that receives considerably less attention.
AI may also help us fight the next pandemic.
Imagine that a completely new virus appears tomorrow.
Once its genome has been sequenced, computational systems can begin analysing its proteins, identifying promising targets and proposing vaccine antigens, antibodies or antiviral compounds at a speed that would have been unimaginable only a few decades ago.
During COVID-19, the genetic sequence of SARS-CoV-2 became available in January 2020. According to CEPI, researchers at the US National Institutes of Health designed a vaccine candidate within two days of receiving that sequence.
Two days.
That does not mean we can produce a vaccine in two days, because computers run considerably faster than clinical trials.
A promising molecule or vaccine design still has to be made, tested, manufactured and shown to be sufficiently safe and effective in human beings. Regulators need evidence. Factories need to produce millions or billions of doses. Distribution has to follow.
COVID vaccines took 326 days from identification of the virus to the first vaccine rollout, an extraordinary achievement at the time.
The international ambition now is even more dramatic.
CEPI’s 100 Days Mission aims to compress the development of vaccines against a newly emerging pandemic threat to roughly one hundred days.
AI could become an important part of that acceleration.
Which creates a peculiar biological arms race.
AI helps us explore what viruses could become.
AI helps us predict how they might evolve.
AI can help design vaccines and therapeutics against them.
Get Sophy Smythe’s stories in your inbox
Join Medium for free to get updates from this writer.
And AI can now generate viral genomes itself.
The technology isn’t standing on either side.
We are.
What if the virus and the cure have the same inventor?
This is where the question becomes much more interesting to me.
Why would an intelligent person deliberately use AI to design a dangerous virus?
The obvious answer is terrorism, warfare or some deranged desire to destroy humanity.
But that answer is almost too easy.
Suppose the intention were exactly the opposite.
Imagine a brilliant scientist convinced that another pandemic is inevitable. Nature constantly mutates viruses without asking permission from ethics committees, governments or pharmaceutical companies. Somewhere, eventually, evolution may produce precisely the combination we fear.
So our imaginary scientist proposes something apparently sensible.
Don’t wait for nature.
Use AI to predict it.
Create millions of possible future viruses digitally. Identify the dangerous ones before they exist. Develop vaccines and treatments against them in advance.
For the first time, humanity might be waiting for the pandemic rather than the pandemic waiting for us.
It sounds brilliant.
Until one of those designs leaves the computer.
Then we have an entirely different problem.
And perhaps the most unsettling question isn’t whether AI could design a dangerous virus.
It is whether another AI could design the defence quickly enough.
The computer might propose a vaccine or antibody within hours or days, but human biology still works on human time. Testing safety takes time. Clinical trials take time. Manufacturing takes time.
A virus does not have those inconveniences.
While researching my latest medical thriller, The Hellraiser, I imagined an AI capable of designing a virus that had never existed before. I pushed the science forward because that is one of the privileges of writing fiction: you take something possible, move it several uncomfortable steps ahead and ask, What if?
I thought I was writing about the future.
Now scientists have used AI to design functioning viruses, other scientists are engineering viruses to fight cancer, and researchers are using AI to accelerate vaccines against the pandemics we haven’t met yet.
None of that means the nightmare in The Hellraiser is real.
Thankfully.
But the distance between “What if?” and “What now?” is becoming considerably shorter.
And I find myself wondering:
If AI can eventually help us design both the virus and the cure, which one will move faster?
#ArtificialIntelligence #GenerativeAI #Biotechnology #FutureOfMedicine #TheHellraiser
