AI-made virus marks a breakthrough for science, but some experts are worriedScientists have used AI to design functioning viruses from scratch in a major biotechnology breakthrough that is dividing researchers.Alex Blair@alexblair_13 min readAugust 7, 2026 - 9:16PMScientists have used artificial intelligence to design functioning viruses from scratch in a major biotechnology breakthrough that researchers say could help fight drug-resistant infections.But the landmark moment has also reopened difficult questions amongst experts about how far synthetic biology should be allowed to go.Researchers at Stanford University used genome language models called Evo 1 and Evo 2 to generate thousands of complete viral genomes.Nearly 300 were selected for laboratory testing and 16 produced viable bacteriophages, which are essentially viruses that infect bacteria rather than humans.A cocktail of the AI-designed phages was then able to kill strains of E. coli that had developed resistance to naturally occurring bacteriophages.The work could have major medical applications, which has always been one of the big selling points for rapid AI advancement.Others you may likenewsnewsThe Stanford team deliberately excluded genetic information from viruses capable of infecting humans, animals and plants from the relevant training data.Researchers also acknowledged the work raised “important biosafety, biocontainment and biosecurity considerations”.Antimicrobial resistance is already associated with millions of deaths globally, and the World Health Organisation (WHO) says phages are being explored as an alternative treatment for bacterial infections that no longer respond to antibiotics.The WHO has not called for a ban on AI-driven biological research, but has repeatedly warned that the collision of AI, synthetic biology and pathogen manipulation creates new dual-use risks that governments must actively manage.Naturally, the breakthrough has sounded huge alarm bells for biosecurity experts.Johns Hopkins researchers Tom Inglesby and Moritz Hanke warned in an accompanying Science article that governments are not ready for the risks it could pose.“Although this is promising for life sciences applications, it also raises urgent biosafety and biosecurity questions. The ability to compose viral genomes using generative AI now exists; the governance to safely steer it does not.”World Health Organisation Director-General Tedros Adhanom Ghebreyesus. The WHO has not called for a ban on AI-driven biological research, but has repeatedly warned that the collision of AI, synthetic biology and pathogen manipulation creates new dual-use risks that governments must actively manage. (Photo by Fabrice COFFRINI / AFP)Is this gain-of-function research?No. Well, not in the sense that usually sits at the centre of contemporary discussion over human-made viruses.The Stanford researchers designed new bacteriophages to attack bacteria rather than modifying a human pathogen to make it more transmissible, virulent or otherwise dangerous.But experts say the fields intersect because both raise the same, obvious problem. There has been some pushback to those fears, with Imperial College London synthetic genome expert Tom Ellis cautioned against overstating the immediate AI threat.“This is literally the smallest and easiest genome to make,” he said via the Guardian.“But honestly, the threat from full AI design and writing of a genome of a virus or bacteria is very overblown when we consider that just taking existing pathogens and making gain-of-function changes to their genomes is so much easier and much more likely to be a real pathogenic threat.”Nevertheless, technology developed for legitimate medical purposes could potentially be adapted for harmful ones.Scientists have been synthesising and modifying viruses for decades, including reconstructing poliovirus and the 1918 influenza virus.The latest development has changed the game in a big way, introducing AI into the process to attempt to help invent the genome itself, rather than scientists merely reproducing or editing an existing genetic blueprint.3d illustration pathogenic viruses causing infection in host organism, Viral disease outbreak, virus abstract background.The technology is still in its infancy, compared to its logical conclusions.In the current research, only 16 of the hundreds of AI-generated genomes tested produced viable viruses, and there is no evidence the same approach can currently generate a functioning human pathogen.‘Unsettling and staggering’The case for AI-made viruses made by advocates is relatively simple. Bacteria continually evolve resistance to both antibiotics and bacteriophages.If scientists can rapidly design new phages to target resistant strains, the technology could dramatically expand the treatment options available when conventional antibiotics fail. The WHO describes phage therapy as promising, although it says considerably more clinical evidence is still needed before it becomes widely available. But the opposing concern is where the technology eventually leads.The primary worry about anything AI-related is that nobody, not even the world’s best minds, can agree upon where this period of rapid advancement ends up.Inglesby and Hanke warned that applying the approach to more dangerous organisms could theoretically produce “new pathogens that … cannot be contained by existing countermeasures”.Inglesby and Hanke warned that applying the approach to more dangerous organisms could theoretically produce ‘new pathogens that … cannot be contained by existing countermeasures’Biosecurity expert Filippa Lentzos from King’s College London argues regulation should therefore extend beyond AI itself and focus on the point where digital genetic instructions become physical DNA.“It’s important to see the bigger governance picture and not focus regulation solely on the AI model,” she said.“A layered approach makes more sense: safeguards around model development and access, responsible research review, synthesis screening, and established laboratory biosafety and biosecurity.” Princeton chemistry professor Michael Hecht, who works on novel proteins and artificial genomes, found the philosophical implications much harder to shrug off.“I find this very unsettling and staggering,” he said via TheWashington Post.He described the arrival of machines generating novel biological designs as “Darwin 2.0.” Tina Hernandez-Boussard, professor of medicine at Stanford has also pushed for immense caution around the work. Her warning centres around the concern that safeguards need to evolve with increasingly capable models.“These models are smart,” she said.“You have to remember that these models are built to have the highest performance, so once they’re given training data, they can override safeguards.”More related storiesMilitary‘Jamming’: US military’s catastrophic failA flight slammed into a mountain near Roswell, New Mexico on a clear night. A report has revealed a clue why, and it’s a problem that’s exploding.Read moreMilitary‘World must see’: Horrifying drone attackDisturbing footage shows an innocent Ukrainian fruit vendor terrorised by a Russian pilot on a “drone safari” hunting for civilians.Read moreInnovation‘No freedom’: Billionaire’s future predictionOpenAI CEO Sam Altman has warned people may be left with no freedom or privacy in a future controlled by artificial intelligence.Read more