A California-based startup says a patient with an antibiotic-resistant infection has received a treatment containing viruses designed by artificial intelligence (AI), in what it describes as a world first.
(Also read: Scientists discover 237-million-year-old fossils of first known amphibian to live entirely in the ocean)
Radical Numerics developed the experimental viruses using its AI model, Omnii-1. The announcement, also reported by The Independent, could mark a significant step towards treating infections that conventional antibiotics struggle to control.
However, the company has not disclosed whether the patient recovered, and its claims have not been verified in a peer-reviewed study.
How AI-designed viruses target bacteria
The treatment involves bacteriophages, commonly called phages, which are viruses that infect and kill bacteria without targeting human cells. Doctors have previously used naturally occurring phages as experimental treatments for difficult bacterial infections.
Finding a suitable phage can be challenging because a virus effective against one bacterial strain may not work against another. Researchers generally have to search collections of naturally occurring viruses to identify one capable of attacking a patient's infection.
Omnii-1 aims to simplify that process. Rather than designing an entirely new virus, the AI model can modify specific regions of an existing phage's genetic material, including those responsible for recognising and attaching to bacterial cells.
{{/usCountry}}Omnii-1 aims to simplify that process. Rather than designing an entirely new virus, the AI model can modify specific regions of an existing phage's genetic material, including those responsible for recognising and attaching to bacterial cells.
{{/usCountry}}According to Radical Numerics' announcement, a mixture containing its AI-designed phages successfully eliminated bacteria isolated from the patient during laboratory testing.
What is known about the patient?
The experimental treatment was subsequently administered under an authorised, doctor-led compassionate-use programme, which allows patients with serious conditions to access investigational therapies when satisfactory alternatives are unavailable.
The startup has withheld details about the patient's condition and treatment outcome, citing privacy concerns.
(Also read: Scientists suggest Van Gogh's Starry Night depicts real sky from July 1889, challenging long-held belief)
It also acknowledged that treating a single patient cannot establish whether the approach is safe or effective for wider medical use.
Earlier AI experiments raise safety concerns
The development follows research announced in August by Stanford University and the Arc Institute, whose scientists used an AI system called Evo 2 to design previously unseen bacteria-killing viruses.
Of nearly 300 viral genomes synthesised and tested, 16 produced functional bacteriophages capable of attacking E. coli bacteria. Those experiments were conducted in laboratories, not on human patients.
While the advances offer potential new approaches to fighting antibiotic resistance, they have also raised biosafety concerns.
As Scientific American reported, researchers have warned that increasingly powerful AI systems capable of designing viruses could be misused, highlighting the need for stronger safeguards as the technology develops.