AI Can Design Bioweapons. Can It Also Stop Them?
The same artificial intelligence that accelerates disease detection can help bad actors engineer deadly pathogens. Public health experts say the race between offense and defense has already begun.
Key points
- AI can compress the time between the first cases of an outbreak and public health detection, potentially preventing a pandemic.
- The same AI tools that help scientists design new biological compounds can, in theory, help bad actors engineer dangerous pathogens.
- Public health experts say early outbreak detection is the single most important factor in stopping a disease from spiralling into a global crisis.
- Speed of response, not the severity of the pathogen, has historically determined whether an outbreak becomes an epidemic.
The question used to be theoretical. Now it is urgent.
Artificial intelligence, the technology behind tools like ChatGPT that can understand language, analyse data and generate new content at speed, is reshaping biology. It can scan millions of genetic sequences, spot patterns in disease data across continents, and help researchers design new medicines in months instead of years. Those are real, measurable benefits.
But the same capabilities cut the other way.
What is the actual threat?
A bad actor with access to AI tools could, according to public health researchers, generate detailed instructions for engineering dangerous pathogens far more easily than before. This is sometimes called "dark biology": the use of AI to design or modify organisms that cause disease, away from any official oversight.
Writing in The Guardian, author Annie Jacobsen frames the moment plainly: this is a race between offense and defense, and both sides are accelerating at the same time.
The concern is not science fiction. Several biosecurity researchers have already demonstrated that current AI systems can lower the technical barrier for someone attempting to misuse biological knowledge. The tools do not need to be perfect to be dangerous. They just need to be good enough.
How does AI help on the defense side?
It helps most by being fast. Every outbreak starts small: a handful of sick people in one place.
Public health wins by catching those cases early, before hospitals fill and containment becomes impossible. Historically, that detection process has been slow. Data sits in local clinic records. Reports travel up bureaucratic chains. By the time an alert reaches the right people, weeks have passed.
AI changes that timeline. Systems trained on hospital admissions, social media posts, pharmacy sales and travel patterns can flag unusual clusters of illness in near real time, giving health authorities days or weeks of extra warning. Days matter enormously when transmission rates are doubling.
Here is a rough picture of what that difference looks like in practice:
| Scenario | Typical detection lag | Outbreak outcome |
|---|---|---|
| Traditional surveillance | 2 to 4 weeks | Often regional spread before response |
| AI-assisted surveillance | 3 to 7 days | Containment more likely |
| No surveillance system | 4 to 8 weeks | High epidemic risk |
Those figures are estimates drawn from historical outbreak analyses, not guarantees. Every outbreak is different.
Should ordinary people be worried?
Worried enough to pay attention, not enough to panic.
The risk is real and the researchers raising it are credible. But the defense side of this equation is also moving fast. Governments, universities and companies are investing in AI-driven surveillance tools precisely because the threat is recognised.
What ordinary people can do is straightforward: stay up to date on vaccinations, pay attention to official public health guidance during outbreaks, and treat sensational claims about engineered diseases with healthy scepticism until confirmed by health authorities.
The honest takeaway here is this: the same technology that makes the threat bigger also makes our early-warning systems sharper. Which side pulls ahead depends largely on how much governments invest in public health infrastructure right now, not after the next crisis starts.
