AI Can Design New Viruses Not Found in Nature. Traders Give Safety Reform Just 10% Odds.

Over the past few years, AI has gotten really good at writing essays, generating funny images, and coding software. It’s taken over search engines, started churning out generic pop songs, and become the corporate world’s go-to customer service option. Now it’s learning to write biology.

According to researchers at Stanford and the Arc Institute, generative AI is now capable of designing brand-new viruses that don’t exist in nature. 

Before you run out into the streets in panic, it’s important to caveat that these AI-generated viruses weren’t designed to kill people. They’re bacteriophages that are specifically designed to target bacteria like E. coli, and the researchers who brought these AI viruses to life in the lab built in loads of safety constraints.

That’s why everybody’s calling this a positive scientific breakthrough. But it also shines new light on just how fast AI can evolve. It’s moved beyond analyzing biological information and has started designing its own functional genomes.

This has huge implications across both business and science. Yet without safety regulations in place, there’s a growing disconnect that we might want to start thinking about.

What Have Scientists Actually Created?

The headline technology behind this breakthrough is called “Evo.” It’s essentially an entire family of AI models that have been co-developed by scientists at Stanford, UC Berkeley, and the Arc Institute. And the concept behind Evo is pretty simple: Instead of learning how words fit together like ChatGPT, Evo learns how genetic sequences fit together.

Researchers asked the models to generate new viral genomes based on well-studied bacteriophages, and they created thousands of candidates. The team then selected 285 of those candidates for lab testing and chemically synthesizing DNA. Of those 285 tests, 16 successfully produced functional bacteriophages that inhibited the growth of E. coli strains.

The scientific upside here could be huge. Antibiotic resistance is one of the biggest issues we’ve been facing in modern medicine, and bacteriophages that target and fight that resistance could revolutionize healthcare. The economic implications could be equally enormous. Drug discovery and industrial biotechnology could be streamlined, saving billions of dollars in research and development.

The concern is that this proves biology is programmable.

The team behind Evo has gone to great lengths to make sure this experiment has been safe. However, there’s an underlying principle of dual-use that should give us all reason to worry. After all, a system that can design a bacteria-killing virus doesn’t inherently get the difference between beneficial biological applications and dangerous ones.

Humans make those decisions, and the regulatory frameworks and international agreements that should be guiding this new cowboy industry are still nowhere to be seen. Over on the prediction markets, nobody’s betting on AI safety rules any time soon. Polymarket is currently showing odds of just 10% that U.S. lawmakers will implement meaningful AI safety legislation before 2027. 

That raises a pretty uncomfortable question for scientists, corporates, and consumers alike: What’s going to happen if biological AI evolves faster than policymakers can agree on its boundaries? 

We’re probably not going to come up with an answer anytime soon. But in the meantime, there are also serious market implications worth bearing in mind.

How Does the AI-Biology Arms Race Affect Markets?

Let’s not kid ourselves: The investment opportunity here is astounding. If the technology behind Evo is applied to drug discovery, it could rapidly reduce the amount of time and money it takes to develop new treatments and bring them to market. Biotechnology companies and Big Pharma could tally up some huge wins.

But the risks associated with those wins could be financially significant, too. 

The prediction markets could be right, and consumers may be left waiting for another 12 months before they get some sort of written protections against the rapid evolution of AI. But even then, regulation will eventually come into direct conflict with these advances. Investors need to start thinking about that regulatory risk and how it affects a company’s overall valuation.

And the same rules apply to infrastructure. AI data centers are already draining regions of water and electricity. If produced at scale, biological AI inevitably creates another layer of demand—whether that’s for synthesis, sequencing, or lab automation. 

Every additional demand and emerging capability creates additional potential for regulatory bottlenecks. That means government policy is going to become a far more important variable for investors trying to price in AI’s future, and so it’s worth treading with caution.

At the end of the day, Stanford’s Evo experiment is simply another example of how AI capability is advancing at phenomenal speeds. Even when their risks are easy to understand, policymakers have historically struggled to regulate new technologies. The risks are a lot more high-stakes when it comes to bioengineering, and so governments and investors have a lot of work ahead of them to ensure the upside doesn't outrun their ability to keep the technology safe.

Only time will tell where we go from here, and prediction markets don’t have much faith in the system. But the clock is definitely ticking.