Artificial Armageddon? AI can now be used to design brand-new VIRUSES

It sounds like the start of a sci–fi film, but scientists have shown that AI can design brand–new infectious viruses the first time. 

Experts at Stanford University in California used ‘Evo’ – an AI tool that creates genomes from scratch. 

Amazingly, the tool was able to create viruses that are able to infect and kill specific bacteria. 

Study author Brian Hie, a professor of computational biology at Stanford University, said the ‘next step is AI–generated life’. 

While the AI viruses are ‘bacteriophages’, meaning they only infect bacteria and not humans, some experts are fearful such technology could spark a new pandemic or come up with a catastrophic new biological weapon. 

Eric Horvitz, computer scientist and chief scientific officer of Microsoft, warns that ‘AI could be misused to engineer biology’. 

‘AI powered protein design is one of the most exciting, fast–paced areas of AI right now, but that speed also raises concerns about potential malevolent uses,’ he said.

‘We must stay proactive, diligent and creative in managing risks.’

In the study, the team used an AI model called Evo, which is akin to ChatGPT, to create new virus genomes – the complete sets of genetic instructions for the organisms.

Just like ChatGPT has been trained on articles, books and text conversations, Evo has been trained on millions of bacteriophage genomes. 

The researchers evaluated thousands of AI–generated sequences before narrowing them down to 302 viable bacteriophages. 

The study showed 16 were capable of hunting down and killing strains of Escherichia coli (E. coli), the common bug that causes illness in humans. 

‘It was quite a surprising result that was really exciting for us, because it shows that this method might potentially be very useful for therapeutics,’ said study co–author Samuel King, bioengineer at Stanford University. 

Because their AI viruses are bacteriophages, they do not infect humans or any other eukaryotes, whether animals, plants or fungi, the team stress. 

But some experts are concerned the technology could be used to develop biological weapons – disease–causing organisms deliberately designed to harm or kill humans.

Jonathan Feldman, a computer science and biology researcher at Georgia Institute of Technology, said there is ‘no sugarcoating the risks’. 

‘We’re nowhere near ready for a world in which artificial intelligence can create a working virus,’ he said in a piece for the Washington Post. 

‘But we need to be, because that’s the world we’re now living in.’  

Craig Venter, biologist and leading genomics expert based in San Diego, said he would have ‘grave concerns’ if someone ‘did this with smallpox or anthrax’. 

‘One area where I urge extreme caution is any viral enhancement research, especially when it’s random so you don’t know what you are getting,’ he told MIT Technology Review. 

In their paper, published as a pre–print in bioRxiv, the Stanford team acknowledge ‘important biosafety considerations’ and emphasize the ‘safeguards inherent to our models’. 

For example, they ran tests to ensure the models couldn’t independently figure out genetic sequences that would make the phages dangerous to humans. 

However, Tina Hernandez–Boussard, a professor of medicine at the Stanford University School of Medicine, said these models are ‘smart’ enough to navigate such hurdles. 

‘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,’ she said. 

In another new study, researchers at Microsoft have revealed AI can design toxic proteins that might might evade safety screening systems in place. 

Their study, published in the journal Science, warned AI tools could be harnessed to generate thousands of synthetic versions of a specific toxin – altering its amino acid sequence while preserving its structure and potentially its function. 

Study author Eric Horvitz, chief scientific officer at Microsoft, warned that there are ‘multiple ways in which AI could be misused to engineer biology’. 

‘AI powered protein design is one of the most exciting, fast–paced areas of AI right now, but that speed also raises concerns about potential malevolent uses,’ he said.

‘We expect these challenges to persist, so there will be a continuing need to identify and address emerging vulnerabilities.’