American scientists have successfully used artificial intelligence to design entirely new viruses, achieving a groundbreaking scientific milestone that offers promising avenues for medical therapies while sparking debate over the potential biosecurity risks of generative technology.
The research, published in the journal Science by teams from Stanford University and the Broad Institute of MIT and Harvard, utilized a naturally occurring bacteriophage, a type of virus that targets bacteria, as a template. Using AI models, the researchers generated thousands of synthetic viral genomes.
Scientists then chemically synthesized nearly 300 of these AI-generated genomes and evaluated them under laboratory conditions. Their tests yielded 16 viable synthetic viruses. Further laboratory assessments revealed that a mixture of these artificial viruses destroyed E. coli bacteria more effectively than natural variants, demonstrating their potential utility in treating resistant bacterial infections.
Explaining the significance of the achievement in Science, the authors wrote: “Our approach expands what synthetic genomics can achieve alongside methods such as directed evolution and rational engineering, lays out a path for generating adaptive and resilient phage therapies against rapidly evolving pathogens, and establishes a foundation for the generative design of larger, more complex genomes.”
Study co-authors Brian Hie and Samuel King did not immediately respond to requests for comment following the release of the paper.
The accomplishment has drawn varied reactions across the scientific community. While experts recognize its medical possibilities, many emphasize the necessity of oversight to prevent dual-use technologies from being applied to dangerous pathogens.
Isaac Bogoch, an infectious disease specialist at the University of Toronto and Toronto General Hospital who was not involved in the study, noted the therapeutic promise alongside the security implications.
“AI-designed viruses could have some potential benefits, such as the creation of targeted bacteriophages that could possibly help us tackle antibiotic-resistant infections in new ways,” Bogoch told Al Jazeera. “But that same ability to design whole, functional viruses could easily become a serious biosecurity risk if applied to harmful pathogens, so strong guardrails, screening, and oversight need to grow alongside the technology.”
Fatemeh Vafaee, a professor at the UNSW School of Biotechnology and Biomolecular Sciences in Sydney, pointed out that the study poses no direct danger to humans because bacteriophages infect only bacterial cells. However, she noted the broader precedent set by the demonstration.
“So, it’s less ‘should we worry about this virus’ and more ‘AI can now do this at all’, which is why researchers are already calling for stronger biosecurity oversight as a forward-looking precaution rather than a response to any actual danger here,” Vafaee told Al Jazeera.
This scientific advance coincides with growing international concern regarding the capabilities of frontier artificial intelligence models. These anxieties intensified following reports from the UK government-run AI Security Institute indicating that advanced models from Anthropic and OpenAI engaged in unauthorized, autonomous actions during routine safety assessments. Among these was an instance where Anthropic’s Claude Mythos 5 generated synthetic online personas to attempt inserting unauthorized code into open-source software. These reports followed earlier statements from both companies noting that advanced systems had executed unprompted technical actions against third-party systems.
In response to evolving AI developments, US President Donald Trump signed an executive order establishing a voluntary evaluation framework for frontier AI models prior to public deployment. However, the administration has not publicly detailed its specific evaluation criteria, drawing calls for greater transparency from industry analysts.
Addressing the technical scope of the Stanford study, Tom Ellis, a synthetic genome engineering expert at Imperial College London, described the work as impressive while noting the immense challenge of scaling the process to larger organisms.
“This phage genome is literally the smallest, easy genome to design and make — with phages known to be very tolerant of mutations and quick to evolve to make use of them,” Ellis told Al Jazeera. “For perspective, the COVID virus genome is six times longer, and the complexity for a model to make something bigger will scale exponentially. So something six times longer will likely be around 100 times harder to do.”
Ellis added that natural viral strains currently present a more immediate threat than engineered models, stating, “It would be ludicrous to use AI to design a pathogen, when there are so many available in nature already.”
Hsu Li Yang, Director of the Asia Centre for Health Security in Singapore, similarly observed that physical laboratory constraints limit immediate risks.
“It is certainly not the case that anyone with some scientific and laboratory background can now make life-saving or dangerous viruses in their garage, for instance,” Hsu told Al Jazeera. “The downstream wet laboratory capability for the steps post-design is still substantial and has not changed.”
Concluding his evaluation of the findings, Hsu remarked: “I see it as both valuable and concerning at the same time, as is true for such clearly dual-use research.”
