Intelligence chiefs warn artificial intelligence could transform biological weapons development

Threat Warnings

Artificial intelligence has the potential to change the stakes on existing and new bioweapons, according to Sir Ken McCallum, the head of the UK intelligence service MI5. Speaking in his annual address on national threats, McCallum highlighted risks from both unskilled groups and well-funded states. Rogue actors working in garage labs could use AI chatbots to compile information and plan experiments with existing pathogens. Meanwhile, well-funded nation states with high-containment labs could use AI tools to find genetic tweaks that make pathogens more lethal. Dr Filippa Lentzos, a security expert at King's College London, notes that AI amplifies existing expertise rather than removing the need for it. A more distant risk involves experts using AI to design entirely new pathogens with no antidotes. Last year, police in India uncovered a suspected bioterrorism plot where a doctor and others allegedly used AI searches to make the deadly toxin ricin with kitchen tools. In September, the AI firm Anthropic reported five cases of scientists trying to use its Claude chatbot for dangerous work, including planning experiments to mutate a bird flu virus to spread more easily in mammals.

Practical Hurdles

Despite these risks, significant barriers remain to actually building biological weapons. Information is only part of the problem, and tools cannot replace tacit unwritten knowhow, access to materials, or proper equipment. Sonia Ben Ouagrham-Gormley of George Mason University points out that tacit knowledge has driven past state bioweapons programs. Professor Martin Paesa at the University of Zurich adds that having plans is useless without physical access to restricted materials like uranium or dangerous pathogens. Experts also warn of a dual-use dilemma where helpful medical research overlaps with weapons development. In 2025, Los Alamos National Laboratory trained an AI to predict how marine snail neurotoxins latch onto cells to help develop antidotes, but Chinese researchers used the work to design new toxins. Professor Tom Inglesby of Johns Hopkins University notes that no country yet has formal government policies to stop AI misuse in this area. He advocates for independent model evaluations before public release and stricter screening by DNA synthesis companies.

Context

Artificial intelligence tools are trained on vast amounts of public data, which can inadvertently include sensitive instructions on chemistry and biology. Security agencies and researchers worldwide are currently debating how to balance open scientific progress with safety controls to prevent dangerous misuse of advanced language models.