Scientists Use AI to Create 16 New Viruses Aimed at Antibiotic-Resistant Bacteria

Scientists have used AI systems to synthesize 16 new bacteriophage viruses, a potential alternative to antibiotics in the fight against bacterial resistance, Wired reports.

What bacteriophages are and why create them

Bacteriophages are viruses that attack bacteria, not human cells. Researchers see them as a promising alternative to antibiotics against pathogens that have developed resistance to treatment. According to the World Health Organization, antibiotic resistance already causes at least 1.27 million deaths a year, and by 2050 that figure could rise substantially, according to forecasts.

Traditionally, phages for therapeutic use were found in nature: in soil, sewage and on plant surfaces. The process is painstaking and unpredictable: the phage you need may simply not exist, or may go undetected for years. Phage therapy, studied since the early 20th century, never reached the mass market largely because of this problem. AI changes the equation: generative models can design new viral proteins in days, exploring molecular structures that scientists could not search by hand even over several lifetimes.

• 16 new viruses created with AI systems
• The goal is to target bacteria that resist antibiotics
• According to the WHO, antibiotic resistance claims 1.27 million lives a year
• Traditional phage hunting takes years; AI synthesis takes days or hours

How does AI create viruses that do not exist in nature?

AI was used to generate new viral protein sequences, a task that is practically impossible to do by hand because of the astronomical number of possible variants. In this sense, AI acts as a molecular designer, generating thousands of candidates where a human team could test only a handful. Importantly, the 16 viruses created are not modifications of known phages: they are synthetic structures that did not previously exist in nature. The resulting organisms are then tested in the lab for their ability to effectively kill target bacteria.

This approach opens the way to personalized phage therapies: it will be possible to design a virus specifically for a particular resistant strain in a particular patient, where no natural counterpart exists or fits. Today, phage therapy is held back by exactly this limitation: matching the right phage on a "key and lock" basis works only in some cases.

"Using AI systems to create viruses opens new possibilities for fighting bacterial resistance," the

Wired piece says.

Why is regulation not keeping pace with AI biotech?

AI-driven virus development is moving faster than the regulatory frameworks around it. This is a familiar problem for biotechnology as a whole, but AI sharpens it: what once required a specialized lab, a large team and many years of work is now achievable much faster and with fewer resources. Barriers to entry are falling, and the potential for abuse is rising with them.

No unified international approach to overseeing AI-synthesized biological entities has been developed yet. Who is entitled to run such experiments? What level of safety review is needed before results are published? The key vulnerability is dual use: the same models that design therapeutic phages could, in theory, be redirected to create dangerous pathogens. That is why biosecurity experts insist that regulation lagging behind AI capabilities is no longer an abstract threat but a growing risk.

What this means

Creating 16 viruses with AI is one of the most concrete examples of how generative technologies are transforming biomedicine: from finding ready-made solutions in nature to designing new ones from scratch. Medicine gains a potentially powerful tool against one of the main crises in modern healthcare. How quickly regulators can close the gap with the lab will largely determine whether this technology becomes primarily a means of treatment or a source of fundamentally new risks.