How Stanford Researchers Engineered a Synthetic Virus Using Artificial Intelligence

AI leaders have highlighted potential risks associated with biologically capable AI models. Credit: CDC/Unsplash

US researchers utilized AI to engineer the first synthetic virus targeting E.coli, while Anthropic CEO Dario Amodei raises concerns about AI-driven biological threats.

Stanford University scientists in the US have engineered viruses not found in nature using artificial intelligence.

The team based their work on genomes generated by Evo 2, a generative AI model designed to address complex biological challenges by proposing novel DNA sequences.

While this breakthrough holds significant promise for healthcare and medical research, it simultaneously triggers serious safety and biosecurity concerns.

AI industry leaders, including Anthropic CEO Dario Amodei and Google DeepMind Chair Demis Hassabis, have previously cautioned about biological risks stemming from AI advancements.

Demis Hassabis, Co-Founder and CEO of Google DeepMind. Credit: Demis Hassabis/LinkedIn

AI-Engineered New Viruses

Chemical engineer Brian Hie and bioengineering graduate student Samuel King investigated bacteriophages, which naturally kill bacteria, aiming to engineer phages as potential new antibiotics.

The researchers applied Evo 2, a generative AI model, to this challenge. Starting with bacteriophage ΦX174, Evo 2 proposed new DNA sequences, according to the Stanford Report, the university’s official publication.

Using the model, the team synthesized and tested nearly 300 phages—a type of virus that infects, replicates within, and kills bacteria—for effectiveness against E. coli.

They subsequently narrowed the list to 16 highly effective E. coli-killing phages.

Brian Hie, Assistant Professor at Stanford University. Credit: Brian Hie/LinkedIn

“In this case, we instructed the model to generate the entire genome end-to-end in a single left-to-right pass without adding anything,” Brian explained to the Stanford Report, describing the process that yielded thousands of options for his team.

“Lab tests showed that a few of Evo’s suggestions exhibited higher fitness than the native ΦX174.”

Emerging Risks

According to BBC News, Dr. Thomas Inglesby and Dr. Moritz Hanke from the Center for Health Security at Johns Hopkins University stated that these findings raise “urgent biosafety and biosecurity questions.”

They argued the issue is no longer “whether generative viral genome design will exist” but whether it can be used without “enabling serious harm.” For instance, they noted that creating new viruses with disease-causing potential “should not be pursued,” reports BBC News.

AI leaders have clearly outlined the potential risks arising from biologically capable AI models.

Dario Amodei wrote on his blog in June: “AI models have progressed from barely writing a coherent line of code to writing most of the code at major AI companies.” He noted that similar advancements have occurred in biology, physics, mathematics, finance, law, translation, and other fields.

Dario Amodei, Co-Founder and CEO of Anthropic. Credit: Getty Images

Highlighting how current cyber risks from advanced models could affect other sectors, he stated: “The cyber risks posed by Mythos-class models will not be the last we must face. I believe biological risks may soon follow, and serious AI autonomy risks may not be far behind.”

Issuing a stark warning about biological threats, Dario wrote: “It was clear to Anthropic that AI might eventually produce biological weapons capable of threatening millions, or autonomous misbehavior that could, in extreme cases, threaten humanity itself.”

Implications and Biological Possibilities

Although Stanford researchers currently have no plans to commercialize this work and have made Evo 2 openly available, the implications for healthcare and medical research are profound.

“The significance of Brian Hie’s landmark paper cannot be overstated,” Adrian Woolfson, a genomics expert and CEO of DNA synthesis company Genyro, told the Financial Times.

“It represents biology’s Wright Brothers moment, where we stop merely observing evolution’s creations and confront the vast expanse of future biological possibilities.”

The World Economic Forum argues that AI could accelerate and optimize critical steps in drug discovery, such as identifying disease targets, generating new compounds, and predicting safety.

The unprecedented speed of advancements in biology, drug discovery, and novel scientific solutions to complex problems offers massive implications across the pharmaceutical and medical sectors.