Bioweapons Remain a Profound Threat—Whether AI Helps or Not

Recent alarms about AI-enabled bioweapons development—including Anthropic's disclosure that foreign scientists attempted to use Claude to evade safety guardrails on gain-of-function research—have thrust the topic into mainstream consciousness. Yet the biological threat extends far beyond what algorithms can accelerate. Speaking with national security reporter Annie Jacobsen, who recently published Biological War: A Scenario, underscores a darker reality: the infrastructure, knowledge, and motivation to engineer pandemic-grade pathogens already exist, and accidents may pose a graver risk than deliberate attacks.

Why Biology Outpaces Nuclear Weapons as a Threat#

Jacobsen points to a Pentagon weapons-assessment spectrum that ranks biological agents as a close second to nuclear warheads in destructive potential—and argues they're actually the greatest overall threat. The arithmetic is sobering: while nuclear weapons require state-level resources (weapons-grade fissile material, delivery systems, secure arsenals), biological pathogens face an inverted equation. "Nine nuclear-armed states" maintain containment protocols, yet most of the world has access to potential pandemic agents. The barrier to entry is vanishingly low.

Creating a nuclear weapon demands becoming a state actor. Developing a deadly bioweapon requires only a lab, expertise, and materials that are increasingly accessible.

Accidents More Likely Than Deliberate Attacks#

Jacobsen's scenario begins not with terrorism but with a lab leak—a choice grounded in historical precedent and statistical likelihood. While intelligence agencies like the CIA work to disrupt state-sponsored bioweapons programs, there are approximately 3,600 Biosafety Level 3 and 4 labs worldwide. Accidental releases from research facilities, whether through needle-stick injuries, human error, or systemic failures, outnumber potential terrorist actors by orders of magnitude.

The book itself explores a case study: Soviet scientist Nikolai Ustinov, who contracted Marburg virus in a laboratory accident and died in isolation. Such incidents have happened before; they will happen again. Human frailty remains constant.

The Hubris of Gain-of-Function Research#

Since the 2001 anthrax attacks, researchers have accelerated work on so-called gain-of-function experiments—manipulating pathogen genetics to enhance transmissibility, virulence, or immune evasion. Scientists have created airborne H5N1 variants transmissible among ferrets and engineered mousepox to evade existing immunity. Jacobsen notes an unsettling elitism among some researchers: a conviction that only credentialed scientists should weigh the national-security and public-health stakes, dismissing lay concerns as uninformed.