Sir Ken McCallum, head of MI5, cautioned in his annual security assessment that artificial intelligence poses risks to bioweapon development by potentially changing the nature of existing threats and creating new ones. Experts identified two primary risk categories: less-skilled actors or terrorist groups operating from minimal facilities could use AI chatbots to compile information and troubleshoot problems related to existing pathogens, while well-resourced nation states with advanced laboratories and trained personnel could leverage AI to identify genetic modifications that enhance pathogen lethality. Dr Filippa Lentzos from King’s College London notes that in the latter scenario, AI functions to amplify existing expertise rather than create it entirely.
Several documented cases have demonstrated attempts to misuse AI for bioweapon-related purposes. Indian authorities uncovered a suspected bioterrorism plot involving queries to AI chatbots about producing ricin, and Anthropic reported five instances of scientists requesting assistance with potentially dangerous viral or toxin research. In one case, researchers allegedly sought help planning experiments to introduce mutations in bird flu virus. Academic exercises, including one at MIT, have shown that non-specialist students can obtain concerning information from chatbots within minimal timeframes, including pathogen identification and DNA synthesis supplier names. However, challenges remain substantial. Creating functional bioweapons requires tacit knowledge, access to biological materials and specialized equipment, experimental validation capabilities, and delivery systems—factors that AI alone cannot provide.
The dual-use nature of biological research complicates oversight, as legitimate pandemic surveillance and therapeutic development can resemble weapon development efforts. In 2025, Los Alamos National Laboratory created AI tools to predict neurotoxin interactions for antidote development, but concerns arose when Chinese researchers adapted similar approaches. Recent achievements in AI-designed viruses that infect bacteria raised additional biosecurity questions despite potential medical applications.
Current safeguards remain inconsistent. Most AI companies have not publicly disclosed bioweapon-specific protections, and no formal government policies exist internationally to address such misuse. Experts advocate for mandatory independent evaluation of AI models before release, restrictions on high-risk biological data, and enhanced screening of genetic material orders by suppliers.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI