In his annual threat address to the nation, Sir Ken McCallum, head of Britain’s intelligence service, MI5, warned that AI had the potential to “change the stakes” on existing and even new biological weapons. Here we examine how AI could contribute to the development of biological weapons and the obstacles that remain.
How could AI help people make biological weapons?
It depends on who wants to make them. One risk is that of a bad actor or rogue, unqualified terrorist group working in a garage laboratory. AI chatbots could help them hone their skills, compiling relevant information, planning experiments and resolving problems. It is likely that work will focus on the use of existing pathogens and toxins.
The other risk comes from well-resourced organizations and nation states that have many trained scientists, high-containment laboratories, and all the equipment and reagents needed to safely work on biological weapons. AI tools could help researchers identify genetic modifications to make existing pathogens more deadly, for example. This is the most immediate concern, according to Dr Filippa Lentzos, professor of science and international security at King’s College London. “Here, AI does not eliminate the need for expertise, it amplifies the expertise already present,” she explains.
A more distant risk comes from highly skilled groups using AI to design completely new pathogens and toxins that are drug-resistant or lack antidotes.
Have people tried to use AI to make biological weapons?
Last year, Indian police uncovered an alleged bioterrorism plot in which a doctor and several others relied on chatbot queries and AI-based searches, allegedly to help them make ricin, a deadly toxin, using kitchen utensils. It’s unclear whether the AI responses helped, but other accounts suggest the tools can provide relevant and dangerous information.
AI company Anthropic raised new concerns in September when it reported five cases of scientists in unnamed countries attempting to use its chatbot Claude for what could have been dangerous work on viruses or toxins. One red flag, the company says, is that after being blocked, scientists found workarounds.
In one case, Anthropic said researchers used the chatbot to plan experiments that would introduce mutations into a highly pathogenic avian flu virus to make it adapt and spread more easily in mammals. This work and other cases could signify attempts to make pathogens more dangerous, but such studies are also done in legitimate labs to help researchers detect future pandemics and develop vaccines and drugs to treat diseases.
What caused concern?
Numerous studies have shown how AI tools can provide potentially dangerous information about biological weapons. In a recent exercise at MIT, non-scientist students were asked if they could get an AI chatbot to help them start a pandemic. Within an hour, the chatbots listed four potential pandemic pathogens, explained how to make them from synthetic DNA, and provided the names of DNA synthesis companies that were unlikely to screen their orders. However, this is easier said than done.
In another case, Dr. David Relman, a researcher at Stanford University, described how an AI explained how to modify a pathogen to make it resistant to treatment. Another scientist revealed that an AI chatbot explained how to use a weather balloon to drop biological weapons from the sky.
The dual-use nature of biological research can make it difficult to distinguish between valuable work that may protect health and attempts to develop biological weapons. In 2025, scientists at Los Alamos National Laboratory trained an AI to predict how exceptionally potent neurotoxins from marine snails latch onto human cells, work that could help develop antidotes. But the American government became concerned when Chinese researchers were inspired by this work to develop their own tool for designing new toxins. The Chinese researchers said their work was aimed solely at developing drugs.
In April, scientists raised new concerns when they created the first AI-designed viruses. It is a special type of virus that only invades bacteria, and this work could lead to new drugs, but experts have warned it raises “urgent biosafety and biosecurity questions”.
What obstacles make things difficult?
Most work on biological weapons is beyond the reach of unqualified groups: in addition to being highly technical, it is extremely dangerous. Although AI chatbots can be asked to provide information about biological weapons, a list of instructions is not enough. “Information is only part of the problem,” Lentzos says. “Tacit expertise, access to biological materials and equipment, experimental validation, scale-up and delivery all remain significant obstacles. »
Sonia Ben Ouagrham-Gormley, deputy director of the biodefense program at George Mason University in Virginia, says that tacit knowledge – the unwritten know-how that an AI is unlikely to help with – has been essential to the success of historic state-run bioweapons programs.
Professor Martin Pačesa from the University of Zurich said access to material was essential. “Any qualified physicist could sketch out plans for a potentially more effective nuclear bomb, but where are they going to get a piece of uranium from? Same for biological weapons. If you have access to the hardware and facilities capable of producing them, then AI won’t make much of a difference if you want to do harm.”
Some AI companies are taking steps to prevent people from using their tools to develop biological weapons, but the majority don’t announce what, if anything, they’re doing, says Professor Tom Inglesby, director of the Johns Hopkins Center for Health Security. “We have yet to have any formal government policy in place in any country to protect against this type of misuse,” he says.
He wants independent formal evaluations of new AI models before they are released, with any hazards eliminated before the models are made public. And high-risk biological data, such as information about how pathogens might spread more easily, should be restricted. Meanwhile, companies that make genetic material for researchers must review orders to ensure they are not aimed at creating new, highly dangerous viruses. Some screenings already take place, but they are largely voluntary and can be avoided if researchers use AI to design dangerous new genetic sequences that companies don’t check.
Gn world
