Scientists have created the first viruses designed by artificial intelligence in a milestone that raises hopes for new medicines and fresh concerns about how to keep the technology safe. The viruses are bacteriophages, which infect only bacteria and are used worldwide to treat patients with persistent infections.

In lab tests, a cocktail of the AI-designed viruses killed E coli bugs that were resistant to natural bacteriophages. Dr Brian Hie, a chemical engineer at Stanford University in California, used genome language models, the genetic equivalent of the large language models behind AI chatbots, to design functioning genomes for bacteriophages. The viruses were then made in the laboratory and tested against E coli in a dish.

The breakthrough was described by the researchers as the first time whole genomes have been successfully designed by AI. They said the work marked a next step in the complexity that can be designed by generative AI, and a potential turning point for synthetic biology.

What did the researchers say the breakthrough could do?

The researchers wrote in the journal that the ability to “rapidly design” genomes and tune them for specific bugs while overcoming resistance could “transform phage therapy” and “expand biotechnological toolkits”.

They also said the work could open the door to new drugs and therapies, including phages to tackle disease, enzymes to treat genetic disorders and antibodies for immunotherapy.

But beyond the potential benefits, the scientists said the work raised “important biosafety, biocontainment and biosecurity considerations” and urged others who were designing whole genomes to “consult both safety and security professionals throughout the project”.

In an accompanying commentary in Science, Prof Tom Inglesby and Dr Moritz Hanke at the Center for Health Security at Johns Hopkins University in Baltimore reinforced the warning, writing:

“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.”

How were the AI-designed viruses made?

Hie and his colleagues used AI models called Evo1 and Evo2 to design the new viral genomes. The models were trained on genetic data from 2 million bacteriophages. The genetic code for viruses that can infect plants, humans or other animals was intentionally excluded from the AI’s training to reduce the risk of it designing dangerous viruses.

The second source said the models were trained on genetic codes from viruses, bacteria, plants and people, before being refined to produce bacteriophages that infect only specific species of bacteria. The researchers said the systems work in a similar way to large language models such as ChatGPT, but predict the language of life rather than words.

The AI generated thousands of potential genomes from which the researchers selected nearly 300 to make in the lab. These were dropped into bacteria, which read the genetic code and churned out the new bacteriophages. The process was not efficient: only 16 bacteriophages proved to be viable, but a cocktail of them swiftly overcame resistance in two different strains of E coli.

The Stanford researchers picked the most promising 302 AI designs and synthesised them in the lab, according to the second source. Of these, 16 proved effective at killing E. coli bacteria. Samuel King, a PhD student in the lab, said the team first realised the phage were working in the early hours of the morning when clear spots appeared on petri dishes growing bacteria.

“We were starting to see these clear spots and it was just extremely exciting,”

he said. When the results were shared with the wider team,

“the room spontaneously burst into applause,”

Hie recalled.

Why are biosafety concerns being raised?

Bacteriophage genomes are tiny, but Inglesby and Hanke said the work nevertheless proved that generative AI could create functioning viral genomes. Whether the same approach could be applied to other viruses was unknown, but they said work on pathogens that could infect humans, animals or plants should not be pursued.

“Such genomes might encode new pathogens that … cannot be contained by existing countermeasures,”

they wrote.

The second source said the same concerns have already been raised that AI-designed viruses could be used maliciously to create new diseases. It added that the Stanford team took steps to maximise safety by excluding viruses that could infect complex organisms from the training database, doing the work on phage rather than viruses that infect people and carrying out the research in a secure laboratory.

How do other scientists view the risks?

Tom Ellis, a professor of synthetic genome engineering at Imperial College London, said the work was impressive, but revealed how hard it would be to make more complex genomes.

“This is literally the smallest and easiest genome to make,”

he said.

An AI trained on the genetic code of dangerous bugs could be used to design more harmful viruses, Ellis said, but controlling access to genetic data and having restrictions on making genomes that look dangerous would help.

“Governments are working hard to do this already,”

he added.

“But honestly,”

he said,

“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.”

Where should regulation focus?

Dr Filippa Lentzos, a reader in science and international security at King’s College London, said the most important point to intervene at the moment was when DNA was being manufactured.

“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.”

What could this mean for future biology?

Prof Marc Güell, from the synthetic biology lab at Pompeu Fabra University in Spain, called the study a

“very significant turning point”

and said that for the first time in history scientists were beginning to design biology on a computer. He said this could help tackle major challenges through phages, enzymes and antibodies designed with AI.

Prof Patrick Cai, chair of synthetic genomics at the Manchester Institute of Biotechnology, said the study was an

“important milestone”

and that its significance extended far beyond phages.

“It suggests that genome language models are beginning to learn the design principles encoded by evolution, opening the door to AI-assisted genome writing,”

he said.

Hie said the work could “massively improve human health” by developing new drugs and therapies. He also said it would probably take a lot of work, but would not be impossible, to attempt some simple organisms, and that the team was “definitely interested in working towards” that.

How far is this from designing living organisms?

Viruses are not alive, and the researchers said it would take another significant leap for AI to generate living organisms. The genetic code of the phage is around 5,400 base pairs long. The smallest genome of a living cell is around 500,000 base pairs, while the human genome is three billion base pairs.

The second source also noted that AI-designed phages are much simpler than living cells, but said the research shows how biology can now be designed on a computer before being built in the lab.

Key Facts

Stanford researchers used AI models Evo1 and Evo2 to design bacteriophage genomes.
Of nearly 300 to 302 designs made in the lab, 16 proved viable and killed E coli.
Experts said the work could advance phage therapy but also raises biosafety and biosecurity concerns.
The AI training excluded viruses that infect plants, humans or other animals from the Stanford dataset.