In an annual address on threats to national security, Sir Ken McCallum, director general of Britain’s domestic intelligence service MI5, warned that advances in artificial intelligence could "change the stakes" around existing and novel biological weapons. His comments form part of a broader debate about how rapidly improving machine-learning tools might affect national and global biosafety.
Security analysts and technical experts say AI could make certain aspects of biological research faster and more accessible. Machine-learning models can assist with tasks such as analysing large datasets, suggesting genetic modifications, or optimising laboratory protocols in silico, and automation could reduce the need for deep manual expertise in routine procedures. Those capabilities, they warn, may be exploited by both state actors and non-state groups to accelerate research or scale up production if other conditions are met.
At the same time, specialists stress that substantial hurdles would remain for anyone seeking to develop and deploy biological weapons. Producing, handling and dispersing hazardous pathogens safely and effectively requires specialised laboratory infrastructure, trained personnel, secure supply chains and robust quality control — elements that are not easily replaced by software alone. Experts also point to biological unpredictability and the complexities of translating computational designs into viable biological agents as significant technical barriers.
Beyond technical limits, observers highlight detection, attribution and response systems that still play a key role in deterrence. Public-health surveillance, international norms and export controls are cited as layers of defense that can mitigate the risks posed by misuse. Policymakers and scientists have increasingly called for closer collaboration between the AI and bioscience communities to identify vulnerabilities, improve safeguards and update regulatory frameworks.
McCallum’s remarks underline growing concern among intelligence and public-health officials about converging technologies. While AI may lower certain hurdles and reshape how biological research is conducted, the consensus among analysts is that the threat landscape will depend as much on access to equipment, expertise and infrastructure as on the capabilities of algorithms themselves — and that governance and international cooperation will be central to managing the risks.