The Next Pandemic May Be an AI One

Since the early days of the pandemic, conspiracy theorists have charged that COVID was a manufactured bioweapon, either deliberately leaked or the result of an inadvertent lab leak. There’s been no evidence to support these speculations, but, alas, that is not to say that such bioweapons aren’t truly an existential threat.  And artificial intelligence (AI) may make the threat even worse.

AI in the lab. Create: Bing

Last week the Department of Defense issued its first ever Biodefense Posture Review.  It “recognizes that expanding biological threats, enabled by advances in life sciences and biotechnology, are among the many growing threats to national security that the U.S. military must address.  It goes on to note: “it is a vital interest of the United States to manage the risk of biological incidents, whether naturally occurring, accidental, or deliberate.” 

"We face an unprecedented number of complex biological threats," said Deborah Rosenblum, Assistant Secretary of Defense for Nuclear, Chemical, and Biological Defense Programs. "This review outlines significant reforms and lays the foundation for a resilient total force that deters the use of bioweapons, rapidly responds to natural outbreaks, and minimizes the global risk of laboratory accidents."

And you were worried we had to depend on the CDC and the NIH, especially now that Dr. Fauci is gone.  Never fear: the DoD is on the case. 

A key recommendation is establishment of – big surprise – a new coordinating body, the Biodefense Council. "The Biodefense Posture Review and the Biodefense Council will further enable the Department to deter biological weapons threats and, if needed, to operate in contaminated environments," said John Plumb, Assistant Secretary of Defense for Space Policy. He adds, "As biological threats become more common and more consequential, the BPR's reforms will advance our efforts not only to support the Joint Force, but also to strengthen collaboration with allies and partners."

Which is scarier: that DoD is planning to operate in “contaminated environments,” or that it expects these threats will become “more common and more consequential.” Welcome to the 21^st century.  

Let's get on all that, then. Credit: DoD

The report specifically calls out Iran, North Korea, the People’s Republic of China (PRC), and Russia as having the knowledge and capabilities for such weapons, and assesses that North Korea and Russia still possess offensive biological weapons (it suspects Iram does too, and notes that China considers biology “a new domain of war”). China and Russia “have also proven adept at manipulating the information space to inhibit attribution, to reduce trust and confidence in countermeasure effectiveness, and potentially to slow decision-making following deliberate use.”

It directs further attention to China: “The United States has compliance concerns with respect to PRC military medical institutions’ toxin research and development given their potential as a biothreat. The PRC has also released plans to make China the global leader in technologies like genetic engineering, precision medicine, and brain sciences.”

Asha M. George, executive director at the Bipartisan Commission on Biodefense, told The Washington Post. “I would not be surprised if by next year they’re saying China has some offensive biological weapons programs. Usually, they just say something like, you are concerned about dual use. And this year they didn’t do that,” He added that Russia remains an equally concerning threat.

Here’s where it gets really scary:

New technologies, such as big data, artificial intelligence, and genomic modification, have the potential to significantly influence the chemical, biological, radiological, and nuclear (CBRN) environment. Such technologies simultaneously offer the prospect for more effective, resilient, and cost-efficient military and civilian solutions while also representing potential new threats from state and non-state actors. The same biological and chemical science advancements created to develop life-saving medical countermeasures could also be used by potential adversaries to develop new or enhanced agents. Technologies intended to reduce testing and production inefficiencies, such as biofoundaries and additive manufacturing, create opportunities to reduce the development footprint and increase the number of proliferation pathways available to malign actors. In this way, emerging and disruptive technologies present both risks and opportunities to the United States, its allies, and partners.

Writing in Vox, Jonas Sandbrink, a biosecurity researcher at the University of Oxford, similarly warns: “large language models (LLMs) like ChatGPT, as well as novel AI-powered biological design tools, may significantly increase the risks from biological weapons and bioterrorism.” AI-powered biological design tools (BDTs), he says, “could allow the design of biological agents with unprecedented properties.”  E.g., ones without any evolutionary constraints or precedents.

Obviously, we need more and better oversight over not just bioweapons but also AI generally, but the fear is that bad actors – whether nations or malign individuals/groups – probably won’t feel constrained by any rules or guidelines such oversight degrees.  Even U.S. Senators, not known for their technical prowess, expressed alarm in a recent hearing.  

One witness at the hearing, Dario Amodei, chief executive of the AI company Anthropic, warned: “certain steps in bioweapons production involve knowledge that can't be found on Google or in textbooks and requires a high level of expertise. We found that today's AI tools can fill in some of these steps.”  He thinks an AI-bioweapon is a “medium-term risk,” and by that he meant: “Whatever we do, it has to happen fast…I would really target 2025, 2026, maybe even some chance of 2024.”

The only thing that has remotely offered me any hope is that, whatever DoD or others are doing, DARPA is already working on it.  It established its Biological Technologies Office (BTO) in 2014, recognizing “the vanishing of once longstanding gaps between the life sciences, engineering, and computing disciplines.” One of the key capabilities DARPA is focusing on “is creating innovative biotechnological approaches to rapidly detect and characterize these threats, preventing surprise and maintaining force readiness.”

Credit: BTO, DARPA

Sure, DARPA is focused on the military, but if its work on the Internet and GPS, among others, ended up with wide-reaching civilian applications, one would hope that its efforts here would as well.  There’s no point to the military surviving a bioweapon attack if all the civilians end up dead.

---------------

With synthetic biology, gene editing, and other biological tools, creating bioweapons is or will soon become much easier, and perhaps much more powerful, than building nuclear weapons. With AI, that will happen much quicker and perhaps become even more dangerous. 

The genie is not going back in the box. We’re not going to unlearn all we now know about manipulating biology. We’re not going to stop using AI. Like all tools, though, they’re neither good nor evil; only how we use them is. Let’s hope we use these right.