C&EN covers a lot of the uses of artificial intelligence in chemistry and biology. Long-term readers will remember that we followed the world of protein structure prediction and the de novo protein design field for years before the 2024 Nobel Prize in Chemistry was announced. So when I got advanced notice of a new paper from Brian Hie and collaborators about designing new bacteriophages, it seemed natural for us to cover it for the innovations therein.
What I did not expect was to open the New York Times and see the same paper being covered on the front page. NPR, Wired, and the BBC covered it. Even the Onion had something to say. What worried me was the framing in several of the articles: mainly, the fear of the potential for AI-built viruses to become bioweapons and of the unchecked use of biological AI by bad actors. In the following weeks, I’ve seen that framing continue to capture the public. But while I acknowledge the risk, I can honestly say it hasn’t been keeping me awake at night.
When I covered the research for C&EN, I asked the authors several questions about their work. We also discussed if the main advance was the ability to design viruses or the ability to design a genome. This is an important distinction. The research team led by Hie is interested in using AI to generate a functional genome, and bacteriophages are the smallest possible thing that requires a genome.
C&EN has several knitters on staff, and we know that the common advice before starting a project is to first test out the combination of yarn and needles by making a small sample square—a swatch. Skip that step and you risk making an entire sweater that looks awful because your needles are the wrong size. These bacteriophages are essentially swatches.
But bacteriophages are also a form of virus, albeit ones that infect bacteria. And synthetic biology is a new and rapidly developing field. The work by Hie’s team marks a potentially significant development for those studying antimicrobial resistance by way of using bacteriophages instead of traditional antibiotics, but those advancements are a long way away. So while I saw this paper and thought “swatch,” other people thought “sweater.”
Most of the “sweater” coverage points out potential biosecurity considerations, and the mainstream news isn’t dreaming up the biosecurity angle out of nowhere. A Perspective article published in conjunction with this paper brought up the concern that the research enterprise does not have safeguards in place against those who may mean to use this technology in harmful ways (Science 2026, DOI: 10.1126/science.aej8512).
It is true that a large framework for regulating AI-designed genomes—for viruses or anything else—does not yet exist. But researchers and ethicists are already organizing themselves and their colleagues. Protein designers, DNA providers, and computational chemists are all having discussions about how to control and screen for the misuse of their technologies. Regulators are also encouraging the building of AI safeguards for research.
It is also worth remembering that right now, the chatbot du jour can’t just spit out a recipe for a bioweapon. The image of a mad scientist in their garage asking an agent to build them the Andromeda strain might be luridly compelling. But the truth is that there are many, much more mundane ways that people with ill intent can choose to do harm.
Every scientific advance has also had its shadow. And we will be watchful for these developments in synthetic biology and AI-powered science. But we are excited by the potential in these fields as well. Of the risks that AI poses to life on the planet, the worthiest of consideration are likely concerns about the water use and emissions of data centers. With good governance, bacteriophages should not be added to that list.
This editorial is the result of collective deliberation in C&EN. For this week’s editorial, the lead contributor is Sarah Braner.
Views expressed are not necessarily those of ACS.
