Artificial Intelligence Used to Design Brand-New Viruses, Raising Scientific and Safety Concerns

LONDON, August 10, 2026: Scientists have used artificial intelligence to design entirely new viruses that were subsequently shown to function in laboratory experiments, marking a significant advance in the use of AI for biological research while raising urgent questions about the safety of increasingly powerful biotechnology tools.

Researchers at Stanford University and the Arc Institute used AI systems to design viral genomes that did not previously exist in nature. The resulting viruses were bacteriophages, a class of viruses that infects bacteria rather than humans. The work represents an important step forward because previous AI applications in biology have generally focused on designing individual proteins, genetic sequences or other biological components rather than complete, functional viral genomes.

The researchers produced hundreds of AI-designed genome candidates for laboratory testing. A small number were successfully converted into functioning viruses, with 16 novel bacteriophages ultimately demonstrated to be capable of infecting bacteria.

Potential medical applications

The research could have important implications for medicine, particularly in the fight against antibiotic-resistant bacterial infections.

Bacteriophages naturally attack bacteria and have long been investigated as an alternative or complement to antibiotics. Scientists hope that AI could eventually make it possible to develop customized phages capable of targeting bacterial strains that have become resistant to existing treatments.

In laboratory experiments, the newly created phages were able to attack strains of E. coli. Researchers believe the ability to design viruses with specific characteristics could eventually expand the possibilities for phage-based therapies.

The breakthrough illustrates how AI is increasingly moving beyond traditional data analysis and into the design of biological systems. Genome-focused AI models can identify patterns in genetic information and generate new sequences, potentially allowing researchers to explore biological possibilities that would be difficult or time-consuming to discover through conventional methods.

A new level of biological design

The significance of the research lies in the fact that the AI-generated designs were not simply copies of naturally occurring viruses.

Instead, the researchers sought to create viral genomes with new genetic sequences while retaining the characteristics necessary for the viruses to function. The successful laboratory results demonstrate that AI models can now contribute to the design of complex biological systems rather than merely predicting how existing systems behave.

Brian Hie, a Stanford researcher involved in the work, described the development as a major step in the complexity of biological systems that can be designed using generative AI.

The achievement nevertheless remains at an early research stage. The viruses created in the study were specifically designed to infect bacteria and were not human pathogens.

Experts warn of biosafety risks

The same capabilities that could accelerate medical research are also prompting concern among biosecurity experts.

As AI becomes better at generating functional biological sequences, researchers and policymakers are increasingly concerned about how such technologies could be misused. Designing a new human virus remains considerably more difficult than creating the bacteriophages demonstrated in this research, and the study does not show that AI can independently create a virus capable of causing a human pandemic.

However, experts say the demonstration represents a meaningful change in the capabilities available to biological researchers.

The concern is not limited to AI itself. A biological design must ultimately be produced and tested using laboratory systems, meaning that oversight of DNA synthesis, laboratory facilities and biological experimentation remains an important part of any safety strategy.

Researchers have therefore called for safeguards to develop alongside AI-driven biological technologies. These could include stronger screening of biological sequences, appropriate laboratory controls and oversight mechanisms capable of keeping pace with rapidly advancing computational tools.

Balancing innovation and responsibility

The development highlights a broader challenge facing the scientific community: technologies capable of producing major medical benefits can also create new risks when their capabilities expand faster than governance systems.

AI-assisted biological design could accelerate the development of new treatments, help researchers study infectious diseases and provide new approaches to bacteria that resist conventional medicines.

At the same time, the ability to generate biological systems that function in the real world makes responsible oversight increasingly important.

Scientists say the immediate achievement should not be interpreted as evidence that AI can create dangerous human viruses on demand. Instead, it demonstrates that generative AI has crossed an important threshold in biological design.

The breakthrough is likely to intensify discussions about how AI models, biotechnology companies, laboratories and regulators should work together to ensure that advances in biological engineering remain focused on beneficial applications.

As AI continues to improve, the central question will be how society can capture its potential to transform medicine while preventing the same technology from creating unacceptable biological risks.