Last year, an AI-designed living virus capable of eliminating E. coli was created. After being reported by the science popularization channel “Pan Science,” it sparked online discussions about whether “Resident Evil” could become a reality. Experts emphasized that the experiment has excluded human pathogenic viruses, focusing on solving antibiotic resistance issues. Currently, there are no biochemical crisis concerns.
After Capcom’s horror game “Resident Evil 9: Village of Shadows” was released, it achieved a record of 320,000 concurrent players on Steam. The return of Leon and the in-game performance continued to generate buzz. This reminded the author of a scientific study related to AI-designed living viruses reported last year by Taiwan’s science popularization channel “Pan Science,” which led netizens to associate it with the biopharmaceutical company “Umbrella Corporation” from Resident Evil.
In September 2025, the prestigious scientific journal “Nature” reported that scientists created the first viruses designed by generative AI, capable of tracking and eliminating E. coli.
Stanford University computational biologist Brian Hie stated that this is the first time an AI system has written coherent genome-scale sequences. The next step will be AI-generated life.
His colleague Samuel King added that designing complete living organisms still requires many experimental advances. But current research shows that AI has potential in designing biological tools and treating bacterial infections. Brian Hie hopes that future strategies can enhance treatments targeting specific pathogens.
How does AI design viruses? The research team used Evo 1 and Evo 2 language models, capable of analyzing and generating sequences, and selected the ΦX174 single-stranded DNA virus containing 5,386 nucleotides as a template. After training on numerous phage genomes, the team further used supervised learning to generate viral genomes that can infect antibiotic-resistant E. coli.
After evaluating thousands of generated sequences, researchers screened 302 viable phages. Experiments confirmed that 16 of these exhibited host specificity for E. coli, and the AI-designed phage combinations could kill three different strains of E. coli, breaking through the limitations of wild-type viruses.
Image source: Research process of Evo 1 and Evo 2 language models designing viruses
After the report by Pan Science, netizens quickly linked it to the antagonist organization Umbrella Corporation from Resident Evil.
They expressed concerns that pharmaceutical companies might embark on a path of self-production and self-sale, joking about investing in Umbrella or asking where they could hire protagonist Leon. Some also worried that human morality might not keep pace with technological development, fearing that before curing diseases, a bioweapon crisis might be triggered first, joking, “How many more episodes can we escape?”
In Resident Evil, Umbrella was founded in 1968, ostensibly as a multinational pharmaceutical company, but secretly conducted T-virus research at the Raccoon City laboratory. In the game’s 1998 storyline, the virus leaked through rats and drinking water, causing the entire city to become infected with the “cannibal disease.” The U.S. federal government ultimately destroyed the city with nuclear bombs, and the company went bankrupt.
Image source: Resident Evil, the antagonist organization Umbrella Corporation
Regarding the risks posed by this technology, Peter Koo, a computational biologist at Cold Spring Harbor Laboratory (CSHL), believes that this research lays the foundation for more daring future applications.
He pointed out that relying solely on models is insufficient to generate viruses, but with filtering and system integration by the team, it has been confirmed that this could be a way to produce functional genomes.
Regarding the ethical concerns of AI designing pathogenic viruses, Kerstin Göpfrich, a biophysicist at Heidelberg University, believes that this dilemma of positive and negative applications is not unique to AI; the biological sciences have always had concerns about whether technology can bring benefits or harm.
The paper also notes that the team has excluded viruses affecting eukaryotic organisms, including humans, from their training data. The ΦX174 phage and E. coli system used are non-pathogenic and have a long history of safe use in molecular biology research.
The research team hopes that this method can safely generate AI-designed viruses to address the growing public health issues like antibiotic resistance. Kerstin Göpfrich remains optimistic about this technology, believing it is a field with growth potential and looking forward to future developments.
Further reading:
Tether ventures into AI health tech! How does QVAC Health integrate biological data while protecting your privacy?
The world’s first bio-computer! Human brain cells integrated with silicon chips—will Cyberpunk 2077 become a reality?