Scientists at Stanford University have used artificial intelligence to design 16 fully functional viruses, in a breakthrough some researchers have called a significant turning point in science.
The researchers used AI to create a genome for a bacteriophage, a virus that infects bacteria. The AI suggested thousands of possible genomes, and the team built 302 of them in the lab before testing them against bacteria. Sixteen of the AI-designed viruses were able to kill E. coli.
According to the researchers, the findings suggest AI can design new viruses with the potential to massively improve human health.
Not everyone is convinced the development is free of risk. Dr Simon Clarke, Associate Professor in Cellular Microbiology at the University of Reading, told the Daily Mail the study poses no direct threat to human health because it only involves viruses that target bacteria. But he said creating novel synthetic life forms that do not exist in nature raises broader ethical concerns.
Clarke said the study confirms AI can design viable, non-natural viruses, and that this underlying capability means the technology could potentially be used to create dangerous pathogens with the potential to become a bioweapon. He added that safeguards within the AI model, which were in place for this study, can help mitigate the risk.
Could AI spark a pandemic?
Asked whether AI could create a virus capable of sparking a new pandemic, Clarke said AI can only design genetic code and cannot physically create or release a virus on its own. He said that while the code itself could be misused, sparking a pandemic would require human intervention, specifically bad actors synthesising and assembling the virus in a lab.
Dr Carl Mayers, a Senior Adviser in Biosecurity Policy at the Centre for Long-Term Resilience, agreed the AI could be misused but said doing so would be extremely hard. He told the Daily Mail that at the moment it would be extremely hard, slow and expensive to use this approach to make a bioweapon, noting the researchers had to design, synthesise and test 302 AI-designed genomes to find just 16 that were functional in a simple bacterium. He said it would be much slower, more difficult and more expensive to do this with animal or human viruses.
Calls for stronger safeguards
Mayers said an AI could not spark a pandemic on its own. He said the greater concern is who has access to these powerful biological AI tools, adding that in the right hands they will bring new drugs and new ways to beat antibiotic-resistant bacteria, while in the wrong hands they could cause damage.
Mayers said decisions need to be made about how these tools are governed, and that safeguards must be put in place quickly, just as with other scientific and medical tools. He said that as labs become more automated in future, effective safeguards must always be in place.
