An artificial intelligence research lab in California said it has administered a custom AI-designed virus into a human patient for the first time as a potential scalable solution to counter antibiotic-resistant superbugs.

AI lab Radical Numerics announced the treatment on Tuesday, but it is pending scientific scrutiny in the form of a peer-reviewed study.

If verified, the treatment, named Omnii-1, could mark the first time AI-designed bacteriophages have been administered to a human patient.

These phages are viruses that specifically infect bacteria. Antibiotic-resistant superbugs are projected to cause nearly 40 million deaths by 2050.

Under current phage treatment approaches for bacterial infections, researchers undertake a time-consuming search of vast environmental collections of phages to find one that happens to match a patient's specific infection.

It starts with clinicians extracting a sample of the bacteria causing a patient's infection to find or design phages that kill the infectious bug.

Key to this challenge is understanding the genetic features of a potential phage that could kill particular bacterial strains.

Phages kill bacteria through a multi-step process in which the virus first binds to specific receptor molecules on the surface of the target bacteria.

They then proceed to enter the bacterial cell to hijack it for their own replication and break it open.

The initial binding of the virus to bacteria is very specific, as a given phage only targets particular strains or species of bacteria.

This specificity is a crucial advantage of phage therapy, which spares beneficial gut microbes that tend to be affected by broad-spectrum antibiotics.

Radical Numerics claims that their AI-driven approach can lead to rapid, on-demand generation of phages that meet patient requirements, accelerating the timeline for treatment.

"In the world's first clinical use of AI-designed phages, designs generated by Omnii-1 were administered to a patient with an antibiotic-resistant bacterial infection under an authorised, physician-led compassionate-use pathway," Radical Numerics wrote in a blog post.

"This milestone marks an early step toward translating AI-guided genome design into personalised therapies," it said.

The startup claimed that its AI model is trained on diverse genomic sequences and phage genomes, as well as data on previously measured phage-host interactions.