AI designs working viruses for first time: What the breakthrough means for healthcare & safety

Artificial intelligence has already transformed drug discovery, protein engineering and medical research. Now, scientists have crossed another major milestone — using AI to design entirely new viruses that worked in laboratory experiments.

The breakthrough is being hailed as a landmark in synthetic biology because it demonstrates that AI can move beyond analysing biological data to actively designing living biological systems.

Researchers say the work could accelerate the development of gene therapies, cancer treatments and new ways to fight antibiotic-resistant bacteria. At the same time, it has reignited debate over biosecurity and the responsible use of increasingly powerful AI tools.

What are AI-designed viruses? 

AI-designed viruses are viral genomes created with the help of machine learning models instead of traditional trial-and-error genetic engineering.

Unlike conventional methods, where scientists modify an existing virus one gene at a time, AI can generate an entire viral genome by learning patterns from millions of DNA sequences.

In the latest breakthrough, researchers designed bacteriophages — viruses that infect bacteria but cannot infect humans, animals or plants. The AI-generated genomes did not exist in nature. Scientists chemically synthesized these genomes in the laboratory and tested whether they behaved like real viruses.

Out of hundreds of AI-generated designs, 16 successfully infected and killed drug-resistant Escherichia coli (E. coli) bacteria, proving that AI-designed viral genomes can produce functional viruses.

How did scientists create them? 

Researchers at Stanford University and the Arc Institute developed the viruses using AI models called Evo 1 and Evo 2, often described as "genome language models."

These models work in much the same way as large language models such as ChatGPT. Instead of predicting the next word in a sentence, they predict DNA sequences by learning the "grammar" of genetic code.

The models were trained on approximately two million bacteriophage genomes, enabling them to recognize patterns that determine how viruses function. Rather than copying existing viruses, the AI proposed entirely new genome sequences that researchers later synthesized and tested in the laboratory.

The study represents one of the first demonstrations that AI-generated viral genomes can function in the real world after laboratory validation.

Why did scientists choose bacteriophages? 

Bacteriophages — or simply phages — are viruses that naturally attack bacteria.

Scientists chose them because they are considered safe for humans and are already being explored as an alternative to antibiotics. As antibiotic resistance becomes a growing global health threat, phages offer a promising way to target harmful bacteria without affecting healthy human cells.

Finding a naturally occurring phage capable of killing a particular bacterial strain can take months. Researchers believe AI could eventually design customized phages within days, making treatments for drug-resistant infections faster and more precise.

Why are researchers designing viruses? 

The objective is not to create dangerous pathogens but to improve medicine.

Scientists are exploring AI-designed viruses for several important applications:

• Gene therapy: Delivering healthy genes to patients with inherited disorders. 
• Cancer treatment: Designing viruses that selectively attack tumour cells while sparing healthy tissue. 
• Vaccines: Developing viral platforms capable of producing stronger and longer-lasting immune responses. 
• Rare diseases: Creating customised viral carriers for conditions that currently have limited treatment options. 
• Synthetic biology: Understanding how viruses evolve and interact with cells to improve future therapies and antiviral drugs. 

Most of these efforts involve modifying harmless or weakened viruses already used in medical research.

Why is AI making such a big difference? 

Traditional viral engineering is slow because researchers must experimentally test thousands of genetic variations before finding one that works.

AI dramatically accelerates this process by evaluating millions of possible genetic combinations in a fraction of the time. It helps scientists identify the most promising designs before laboratory testing begins, reducing both development time and cost.

Rather than replacing scientists, AI serves as a powerful design tool that narrows down the vast number of possible genetic sequences.

What makes this breakthrough significant? 

The achievement marks an important shift in biotechnology. For years, AI has been used to predict protein structures and analyse biological data. Designing an entirely new viral genome that functions in laboratory experiments moves AI into a new role — as a biological designer.

The work suggests AI could become as transformative for virology as it has already become for protein engineering and drug discovery.

Are these viruses dangerous? Not necessarily.

The viruses created in this study are bacteriophages, which naturally infect bacteria rather than humans. They were developed under controlled laboratory conditions and for research purposes.

Scientists emphasize that every AI-generated design requires extensive laboratory testing, validation and safety assessment before it could ever be considered for medical applications. The research does not mean AI can suddenly create dangerous human viruses.

Why are scientists concerned? 

While the study focused on harmless bacteriophages, experts say the technology highlights broader biosecurity questions.

As AI systems become more capable, they could theoretically lower the technical barriers for designing biological organisms if used irresponsibly. Although today's models remain far from designing complex human pathogens, many researchers argue that governance frameworks should evolve before the technology becomes significantly more powerful.

Several experts believe regulating DNA synthesis companies, strengthening laboratory biosafety protocols and carefully controlling access to advanced biological AI models may prove more effective than regulating AI software alone.

Many scientific journals and research institutions already conduct dual-use risk assessments before publishing or supporting research involving advanced biological engineering.

Could this transform healthcare? 

Potentially, yes. If successfully translated into clinical medicine, AI-designed viruses could accelerate the development of gene therapies, improve targeted cancer treatments and make personalized medicines more accessible.

The technology could also shorten vaccine development timelines during future outbreaks and speed up the creation of phage therapies to combat antibiotic-resistant bacteria. Scientists expect AI to become a routine tool in virology, much as it has become indispensable in protein engineering.

Future research will likely focus on making viral therapies safer, more effective and easier to manufacture. At the same time, governments, regulators and biosecurity experts are expected to develop stronger frameworks to ensure these advances are used responsibly.