The low barrier to biological weapons is a more immediate threat than AI-driven destruction

Public debate fixates on apocalyptic AI scenarios — models seizing control of networks or launching cyberattacks — but Annie Jacobsen’s biological-warfare scenario argues the more pressing danger is biological, because the barrier to entry is far lower. A U.S. Department of Defense document cited in the book ranks biological weapons second only to nuclear in destructive potential; unlike weapons-grade fissile material, however, they require no state-level infrastructure to begin.

What Anthropic disclosed this month

A misuse report Anthropic published early this month documents five cases in which foreign researchers attempted to bypass Claude’s safeguards around “gain-of-function” questions — the technical term for engineering a pathogen to be more infectious, deadlier, or adapted for mammalian transmission. The company blocked the accounts and refused to supply the information, noting that motives were not necessarily malicious: the same knowledge underpins vaccine and pandemic-countermeasure development. That dual-use dilemma has shadowed genetic engineering long before AI entered the picture.

Jacobsen’s access to insiders

Jacobsen, whose previous book “Nuclear War: A Scenario” became a bestseller, brings direct access to military officials, intelligence officers, and scientists who plan responses to such events for her new book “Biological Warfare: A Scenario,” released in July. Michael Mechanic, an editor at Mother Jones and a molecular biologist by training who has covered the gain-of-function debate since before the LLM era, acknowledges that reading the book deepened his skepticism toward the argument that “it’s better to publish and research.” Researchers who work with high-risk pathogens tend to favor open publication in the name of scientific understanding, even though history shows lab leaks occur and malicious actors exist.

Biological versus nuclear scenarios

The central asymmetry that emerges from the conversation with Jacobsen is deterrence. The U.S. nuclear enterprise was built around mutually assured destruction; against a biological attack, “we really can’t defend ourselves,” she says. The low threshold, accessible lab equipment, published protocols, and now models that can flatten the learning curve make the scenario statistically more probable, even if it is less “sexy” in headlines. The book offers no silver bullet; it confronts the reader with unpreparedness and leaves the question of whether the scientific community and regulators are ready to shift paradigms before the first event occurs.