They Created New Viruses in a Lab

Stanford and the Arc Institute used AI to write sixteen viruses that have never existed. I read the headline and thought about 2020. Then I looked at the calendar, and it got worse.

I couldn't believe my eyes when I saw the headline. Scientists used AI to create viruses not found in nature. Not modeled. Not simulated. Not predicted. Created. Sixteen of them, in a dish, doing the thing viruses do.

And all I could think about was 2020. About how plausible it is that a virus walked out of a lab six years ago — not on purpose, nobody serious is saying on purpose, just *walked out*, the way things walk out of buildings — and shut down the entire world. Every school. Every restaurant. Your job, my job, the funerals nobody got to attend. All of it rearranged by something the size of nothing.

And how fast we filed that away.

So here's what actually happened, because it matters, and most of the coverage is either screaming at you or patting you on the head.

So both are true, and I can hold both. What I couldn't hold was what happened when I went looking to prove my own first instinct — that we'd forgotten 2020.

We hadn't. We were arguing about it *that exact week*.

On July 29, Anthony Fauci sat in front of Rand Paul's Senate committee and invoked the Fifth Amendment, over and over, for hours. Because Republicans are still actively trying to put him in prison — largely over whether he lied about funding gain-of-function work in Wuhan. Biden pardoned him preemptively for everything from 2014 to 2025. Paul's answer is that a pardon doesn't cover a word he says *now*, and he's pushing a vote to hold him in contempt.

On July 28 — the day before — the White House released the U.S. Government Policy for Stopping High-Risk Life Sciences Research. The rulebook six years of this fight finally produced. The first government-wide, enforceable ban on dangerous gain-of-function research.

On August 6, the paper came out.

Nine days. We finished writing the rules for the last one, and nine days later the field demonstrated something the rules don't have a word for.

“We spent six years failing to establish where one virus came from. This week we learned how to build ones that came from nowhere.”

It bans research that "enhances a property or properties of a biological agent in ways that make it more dangerous" — the line every law firm now walking universities through compliance is quoting back to them.

Read that again. *Enhances.* Every prohibited outcome in that document is a verb you perform on something that already exists — enhance, increase, alter, confer, disrupt.

The one clause about *making* something covers "generating or reconstituting an eradicated or extinct biological agent," which means bringing smallpox back. It does not mean writing something that was never here in the first place.

This paper didn't enhance anything. It didn't modify a pathogen. It composed one.

And I want to be fair to these scientists, because they did the responsible thing, and a bacteriophage study almost certainly isn't what that policy is aiming at. That's not my complaint.

My complaint is that the category itself is aimed one war back. The interagency group that's supposed to be watching AI and biology? Hasn't been convened. Agencies don't have to publish implementation guidance until mid-November. Institutions have until January.

And here's the part I keep coming back to, the thing that actually put a chill in me.

Six years later, we still don't know where COVID came from. The FBI says lab, moderate confidence. The CIA says lab, low confidence. Most scientists still say the market. The WHO's 2025 review said plainly that most of what it needed to settle the question was never handed over.

That isn't a footnote to the story. That *is* the story.

We could not establish the origin of one virus that came from somewhere. We had six years, the full weight of the American intelligence community, and the entire scientific press, and the honest answer is still *split*.

Now we can make them come from nowhere.

I'm not scared of the sixteen phages in that dish. They're a good piece of science and they might genuinely help with antibiotic-resistant infections, which kill people every day right now, today, while we argue. I'm scared of the sentence I just had to write to describe them — that a machine wrote a living thing's instructions and the instructions worked — and of the fact that the only reason it pointed somewhere harmless is that the people at the keyboard chose to point it there.

We got lucky with who went first. Luck is not a policy.

The lesson of 2020 was never really about a lab in Wuhan. It was that we could not find out.

•The science here is careful, and the researchers built in the only meaningful safeguard in the story — voluntarily.

•The rulebook we just finished, nine days before this landed, is written entirely in the language of modifying what already exists.

•Nobody has to convene the AI-and-biology group until later. Nobody has to publish guidance until November. Nobody has to do anything until January.

•The next lab does not have to be careful. That is the whole exposure.