Anthropic’s recent misuse report detailed five cases where foreign scientists tried to evade Claude’s guardrails on “gain of function” inquiries. That’s a euphemism for tinkering with a pathogen’s genes to make it more infectious, deadlier, or better able to spread among mammals. The company denied those users the answers they sought and cut off their accounts, even while noting their motives may not have been sinister. Genetic engineering can be used for both good and evil, a dual-use problem that predates AI.
Annie Jacobsen’s new book, Biological War: A Scenario, follows her earlier work on nuclear war. She has decades of high-level national security reporting behind her, and she speaks with current and former military and government officials about what a biological attack might look like. The book came out in late July.
Gain of Function and the Dual Use Problem
Researchers who work on high-risk pathogens often default to publishing their findings. They argue it’s better to understand which genetic traits make organisms a threat to humankind. That ignores the catastrophic consequences of an accidental release, or a deliberate one.
Jacobsen’s book raises the same question: is humanity really better off doing the work? The writer’s honest reaction is unease. The interview points to real risks, not hypothetical ones.
The Comparison Between Nuclear and Biological Weapons
In her book, Jacobsen reproduces a Defense Department document that ranks weapons of mass destruction by threat. Nuclear weapons sit at the far end of the curve because they are the most destructive. But biological weapons are a close second, and she believes they are the greatest overall threat because the threshold for entry is extremely low.
To have a nuclear weapon, you need:
- Weapons-grade fissile material
- An ICBM
- A B-2 bomber
- Or a nuclear-armed submarine
You have to be a state. A terrorist with a dirty bomb is a different threat entirely.
Who Has Access to Pandemic Agents
The key difference is who holds the means. Nine nuclear-armed states contain their arsenals. The rest of the world holds pandemic agents. Most everyone has access to them.
That becomes terrifying. If you were a bioweapons designer, what would you look for in a pathogen? Jacobsen frames it bluntly: an airborne agent, and a public that doesn’t fully grasp the threat.
Accidental Release vs. Deliberate Attack
Jacobsen’s scenario involves the accidental release of a weaponized pathogen from a high-security lab in Russia. The experts she spoke with position accidental release as a much greater threat than someone deploying a bioweapon.
There are serious counterproliferation efforts at the CIA and other intelligence agencies to disrupt bioweapons programs. Biological weapons programs in the hands of terrorists have been shut down. But the people willing to deploy a biological weapon into a community, knowing the blowback would create a pandemic around the world, are rare.
Many terrorists fit that category, but that’s a smaller group than the people who could accidentally release a pathogen from a lab. The experts she spoke with put the risk in that order.
The Unfinished Thought
The interview text is incomplete at the end, with the final thought about all the people—scientists, nation states, academic institutions—that are manipulating pandemic agents left unfinished. The paper notes that the conversation cuts off mid-thought.
What We Know
- Anthropic’s misuse report documented five cases of foreign scientists evading Claude’s guardrails on gain-of-function inquiries.
- Anthropic denied those users the answers they sought and cut off their accounts.
- Jacobsen’s book, Biological War: A Scenario, was released in late July.
- The Defense Department document ranks nuclear weapons as the most destructive, with biological weapons a close second.
- The CIA and other intelligence agencies run counterproliferation efforts against bioweapons programs.
The takeaway is sobering. The technology exists. The access is wide. The accidental route is the likelier one, and the interview presents that expert assessment.
Source material: “There Are Plenty of Reasons to Be Concerned About Bioweapons Development—Even Without AI,” WIRED.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.

