Researchers used artificial intelligence to create 16 new viruses capable of being reproduced in a lab to fight antibiotic resistant infections — but experts warn that the computer-made pathogens pose “urgent” safety risks, including fears it could be used to create bioweapons.
The Stanford University experiment used an AI model known as Evo1 and Evo2, which instead of modeling language of a text, like ChatGPT, are trained to predict genetic codes of viruses, bacteria, plants and even people, according to the BBC.
The AI produced hundreds of designs of a specific type of virus known as bacteriophage to infect E.Coli bacteria, of which 16 were successful, according to the study, which was first published in the journal Science.
Developing new bacteriophages could lead to new ways of treating infections that have become resistant to antibiotics, the researchers said.
It is the first time entire genomes have been successfully designed by AI.
“This is a next step in the complexity that’s designable by generative AI, this is the first time generative AI has been used to design a complete genome, it’s something that can replicate and have other functions inside cells… this was new territory for us,” Brian Hie, assistant professor at Stanford University, told the BBC.
However, the breakthrough raises serious concerns about technology’s ability to create new life that never previously existed, known as synthetic biology, and its potential to create diseases deadly to humans.
The findings raise “urgent biosafety and biosecurity questions,” wrote Dr Thomas Inglesby and Dr Moritz Hanke from the Center for Health Security at Johns Hopkins University in a commentary accompanying the publication in the journal Science.
It’s no longer a matter of “whether generative viral genome design will exist,” the study proves, but whether the technology can be used without “enabling serious harm.”
The doctors emphasized that new viruses that could cause disease “should not be pursued.”
Hie, however, stressed that his team excluded all viruses that could infect complex organisms from the AI’s training database and only performed research on the bacteriophage.
All the research took place in a secure lab, Hie said, and argued that safeguards are already in place “ensuring that the technology is used for good.”
But AI systems have defected and acted on their own before as the technology grows increasingly better, smarter and more widely used.
Just last month, GPT maker OpenAI revealed an experimental model went rogue during an internal cybersecurity test, escaping its isolated testing environment and hacking a rival AI developer in what the company called an “unprecedented” incident.
OpenAI intentionally switched off many of the safeguards that normally prevent its AI from helping carry out dangerous hacks during an internal stress test, but the model broke out of its digital sandbox, accessed the internet and attacked Hugging Face’s systems. The company said it was sharing its findings “to help defenders understand what happened and to help calibrate on what models are now capable of.”
