Biological Weapons in the Age of AI: Why Experts Fear Disease Could Be Engineered

In October 2026, Ken McCallum, the head of Britain's domestic intelligence service MI5, issued a warning that belongs to a category of threat once confined to thrillers: artificial intelligence, he said, could raise the risk posed by existing or new biological weapons, including in the hands of non-state actors. It is the kind of statement that has pushed scientists, security services and international organizations to ask how such a scenario would be managed.

The worry rests on two parallel developments. Advances in biotechnology allow pathogens to be analyzed and altered with growing precision. AI, meanwhile, can process enormous quantities of data at remarkable speed. Experts caution that, over the long term, the two together could be misused to develop biological weapons.

A report from the American AI company Anthropic, published in September 2026, gave the debate a concrete anchor. The company said it had identified and blocked numerous attempts to use its assistant, Claude, in connection with possible biological weapons research. Those attempts included grant applications for so-called gain-of-function research, and projects looking into variants of avian influenza that might be the most transmissible to mammals. Anthropic also stressed that such work, though it could have military uses, can equally serve medical prevention. That ambiguity is among the hardest problems in controlling modern biotechnology.

Biological weapons themselves are old news. Bacteria, viruses, fungi and certain biological toxins can all be weaponized. Plague, smallpox and botulinum toxin are familiar names, and anthrax is among the most notorious, its resistant spores able to survive long periods in the environment and the subject of state weapons programs. The World Health Organization and the UN Office for Disarmament Affairs define a biological weapon as the deliberate release of such substances to harm people, animals or plants. Experts typically break one down into an agent, such as a virus or bacterium, and a delivery mechanism, such as missiles, spraying systems or contaminated food. Alongside nuclear and chemical arms, they count among the ABC weapons and are often classed as weapons of mass destruction.

They differ from chemical weapons in how they act. Agents like sarin or mustard gas injure through toxicity, often within minutes or seconds; examples include chlorine and mustard gas in World War I, and sarin in the Syrian civil war and under Saddam Hussein in Iraq. Biological weapons cause infectious disease, with effects that surface hours, days or weeks later. Some pathogens spread from person to person, which makes the outcome especially hard to predict.

The law is clear on paper. The Biological Weapons Convention came into force in 1975, banning the development, stockpiling, acquisition, retention, production and transfer of biological and toxin weapons. A product of the earlier Geneva Protocol of 1925, which banned their use in warfare, it has been ratified by 189 countries. Israel, Egypt and Syria are among the few that have not signed or have signed without ratifying. Members commit to destroying stocks, production sites and means of delivery.

Enforcement is the weak point. Chemical weapons have the Organization for the Prohibition of Chemical Weapons, which can inspect and oversee the destruction of stockpiles. Nothing comparable exists for biological weapons, because states fear inspections could expose sensitive medical, biotech or pharmaceutical data. Control therefore depends largely on transparency, voluntary reporting and good faith. The limits of a ban alone were shown in the United States in 2001, when letters containing anthrax spores arrived a few weeks after the September 11 attacks. Twenty-two people were infected and five died.

Detection is harder still. Much of the equipment used to make vaccines and medicines could in theory serve military research, the dual-use problem. And an attack can look like a natural outbreak, with deliberate release established only after extensive investigation. History offers grim precedents, from poisoned wells to plague victims' corpses catapulted over besieged walls, though how effective those methods were is disputed. What to watch now is whether verification can evolve as AI and biotechnology make the line between science and misuse harder to draw.