AI helps scientists design novel viruses, raising new questions about biological security

Yesterday 11:18

By: Sahili Aya

AI helps scientists design novel viruses, raising new questions about biological security

Artificial intelligence has taken a new step into the field of synthetic biology, with US researchers using an AI system to generate previously unseen variants of a harmless virus.

The research, published in Science on August 6, demonstrates how AI can analyze enormous quantities of genetic information and use patterns found in biological sequences to propose new viral designs.

The achievement could open new avenues for biomedical research, but it is also raising concerns about how increasingly powerful AI systems might eventually be used to design dangerous biological agents.

AI analyzes billions of genetic patterns

Researchers based in Palo Alto, California, used an AI system called Evo, which is designed to process biological sequences in a way broadly comparable to how language-based AI systems analyze text.

The system was trained using an enormous genetic dataset containing approximately 9 trillion sequences derived from millions of animals, plants, microorganisms and viruses.

To reduce potential risks, the researchers excluded sequences associated with viruses known to infect humans, animals, plants and fungi.

The team then focused the experiment on Phi X-174, a virus that infects E. coli bacteria but is not considered a threat to human health.

Hundreds of thousands of designs generated

Using the genetic information it had analyzed, Evo generated approximately 700,000 potential versions of the selected virus.

Researchers subsequently identified 285 candidates for experimental testing. Ultimately, 16 produced viable viruses in laboratory experiments.

The resulting viruses remained closely related to the original organism and showed comparable levels of resistance, according to the research described.

The findings demonstrate that AI can move beyond simply analyzing existing biological information and can instead generate new genetic possibilities based on patterns learned from large datasets.

Potential benefits for medicine

The researchers believe the approach could eventually contribute to scientific and medical research.

AI systems capable of understanding biological sequences could help scientists explore how genetic changes affect viruses and other organisms, potentially providing new tools for biotechnology and therapeutic development.

However, the technology remains at an early stage, and further experiments will be required to determine whether the approach can be reliably applied to other biological systems.

Security concerns accompany the breakthrough

The same capabilities that make AI potentially valuable for biomedical research also raise questions about biological security.

Researchers and policymakers are increasingly examining the possibility that advanced AI could eventually be misused to design harmful pathogens. The concern is particularly significant as AI systems become better at interpreting and generating biological information.

The current experiment does not involve a virus capable of infecting humans, and the researchers deliberately excluded dangerous viral sequences from their analysis. Nevertheless, the work illustrates how rapidly AI is becoming integrated into biological research.

The development is likely to intensify calls for clear rules governing the use of AI in biotechnology, particularly as researchers continue testing whether similar systems can be applied to a wider range of biological questions.