The Four Pillars: A Hypothesis for Countering Catastrophic Biological Risk

Biological risks are more severe than has been widely appreciated. Recent discussions of mirror bacteria highlight an extreme scenario: a single organism that could infect and kill humans, plants, and animals, exhibits environmental persistence in soil or dust, and might be capable of spreading worldwide within several months. In the worst-case scenario, this could pose an existential risk to humanity, especially if the responses/countermeasures were inadequate.

Less severe pandemic pathogens could still cause hundreds of millions (or billions) of casualties if they were engineered to cause harm. Preventing such catastrophes should be a top priority for humanity. However, if prevention fails, it would also be prudent to have a backup plan.

One way of doing this would be to enumerate the types of pathogens that might be threatening (e.g. viruses, bacteria, fungi, etc), enumerate the subtypes (e.g. adenoviruses, coronaviruses, paramyxoviruses, etc), analyze the degree of risk posed by each subtype, and then develop specialized medical countermeasures in a prioritized way in advance of a threat.

This approach is a good way of tackling naturally occurring threats, but seems ultimately doomed against a determined adversary. It would be hard to feel confident that any list of threats was fully exhaustive—for example, mirror bacteria would have likely been left off of most lists like this. Second, even for threats on the list, surprising breakthroughs or concerted efforts could break assumptions around how effective a given medical countermeasure is. The public history of the Soviet bioweapons program^[1] gives ample examples: attempts to develop hybrids of smallpox and Ebola, engineering a strain of plague resistant to over 15 kinds of antibiotics, and modifying pathogens to intentionally trigger autoimmune reactions. There are simply too many ways that a creative adversary could cause catastrophic harm with a biological weapon, and we need fully generalized defenses that work in a future-proofed way.

Here we outline a hypothesis for ‘four pillars’ of biodefense that should work against even the most sophisticated engineered pathogens or ‘unknown unknowns’. These are:

1. Personal protective equipment (‘PPE’)
2. Pervasive physical barriers and layers of sterilization (‘biohardening’)
3. Pathogen-agnostic early-warning systems (‘detection’)
4. Rapid, reactive medical countermeasures (‘MCMs’)

The first three pillars are our current best guess for defenses that would provide widespread protection and keep society running. And they are robustly ‘future-proof,’ since they exploit fundamental constraints that all pathogens must face. Despite a potentially vast space of possible biological attacks, the problem can be dramatically simplified if we notice that any pathogen will inherently need to first physically enter a human body in order to cause harm.^[2] Similarly, a pathogen that could cause catastrophe must spread widely and produce harmful biological effects, making it detectable.

The first three pillars of the plan are targeted at the earliest stages of a catastrophe, with the goal of saving as many lives as possible and preserving industrial and scientific capacity for the rest of the world to respond more fully. For example, PPE can protect essential workers early on in a catastrophe, which keeps critical industries running—buying time to create even more protective equipment and novel countermeasures. Targeting this leveraged period of time means that achieving widespread protection under the three pillars might be feasible by the end of 2027 with a budget of less than $1 billion.

The fourth and final pillar of the plan is to ensure medical countermeasures can be created after the threat is identified and characterized. There is an open question about whether medical countermeasures could work against any possible biological weapon in the future—my hunch is that in theory they ought to^[3]—but getting into that is beyond the scope of this post. This post focuses on the first three pillars, which are especially neglected and tractable.

PPE

Personal protective equipment (PPE) is a straightforward way to prevent pathogens from entering a human body. PPE comes in many varieties and intensities, ranging from surgical masks and latex gloves to full HAZMAT suits with independent oxygen supplies. One could imagine an extreme hypothetical in which everybody had access to a HAZMAT suit—in such a hypothetical world where people were wearing them, it would be very difficult for any biological weapon to cause much damage even if it was highly contagious and lethal.

What kind of PPE is most needed? If we enumerate every physical pathway by which a pathogen could enter a human body (ingestion, inhalation, skin breaks, etc), we find that the respiratory path is the weakest link. There are good first principles reasons for this: the surface area of our lungs is large compared to other surfaces, we can’t go long periods of time without breathing, it is less difficult to sterilize and/or protect alternative pathways like water, food, surfaces, skin, etc). It also matches what we see empirically: respiratory diseases still cause pandemics in wealthy countries, whereas public health measures have eliminated or drastically reduced other transmission pathways like water-borne disease.^[4]

Here are five types of PPE that protect against inhaling pathogens:

Our current assessment is that elastomeric respirators occupy the ‘sweet spot’ between degree of protection and cost. Each of these masks typically offers a protection factor exceeding 100 (meaning they filter out 99% of incoming particles), costs less than $50, can be reused for six months without needing to change the filters,^[10] and can last more than 10 years in storage. We also think a dedicated effort could reduce the cost substantially—some overseas manufacturers already claim to make them for below $5.^[11]

Although it’s uncertain, different studies suggest that protecting between 12-45 million people in the US would likely be sufficient to prevent infrastructure collapse—roughly the number of workers who would need to go outside of their homes and perform critical tasks like maintaining power plants, water sanitation, hospitals, etc while everybody else stays at home to wait for more permanent solutions. Stockpiling 40 million respirators at $50 each would cost less than a single B-2 bomber, and any industrialized country could obtain similarly cost-effective stockpiles.^[12] If the cost could be reduced further, even independent philanthropists could potentially pay for enough respirators to protect large fractions of the world’s critical workers.

Also, a protection factor of 100 might be sufficient to block all human-to-human repository transmission—even against future ‘unknown unknowns’—so long as the mask provides source control (i.e. filtering exhaled air as well as inhaled air). If each person both exhales and inhales through a mask with a factor of 100, this provides a combined protective factor of 10,000 (meaning a reduction of 99.99%). Transmissibility is upper-bounded by human respiratory aerosol production, and a 10,000-fold reduction might be sufficient to prevent transmission even under worst-case theoretical limits.^[13]

One key assumption is that people are motivated to wear the equipment properly. Our view is that in a truly catastrophic scenario—where infection meant certain death—motivation would be sufficiently high for people to use the equipment to their best ability.

Another key assumption is that there are plenty of safe pathogen-free areas where people don’t need to wear the equipment as they eat, sleep, and so on. In most pandemic scenarios that rely on human-to-human transmission, this could be achieved straightforwardly with isolation and social distancing. In a mirror bacteria scenario though, there could be widespread environmental contamination and environment-to-human transmission, which would require distinct efforts to ensure the presence of safe areas where people could work and live (see the next pillar, ‘biohardening’).

With the mirror bacteria scenario, we’re less confident that a protective factor of 100 would be sufficient to protect people working outside (since the exposure is potentially coming from the environment rather than a source that can also be covered with a mask). On the margin, this would be a reason to stockpile elastomeric respirators that have even higher protective factors, although there would likely be a tradeoff with cost and abundance. Some elastomerics regularly achieve protective factors exceeding 300 or even 1,000, and some military-grade gas masks can achieve protective factors exceeding 10,000. Even if the elastomerics were insufficient on their own, they could be supplemented with additional layers of defense—for example, through the use of improvised powered air purifying respirators (PAPRs) that could fit over the elastomerics.^[14]

Tentative victory condition: most of the world’s critical workers have access to an elastomeric respirator (rated fit factor >100) within 24 hours. Achieving this just for the US could potentially be done by the end of 2027 with a concerted effort aimed at stockpiling 30 million respirators with less than $500M, and other countries could be protected with comparable speed and per capita cost.

Biohardening

Even with sufficient PPE, we will need pathogen-free spaces in which to eat, sleep, work, and live. Achieving this could be relatively straightforward in a strictly human-to-human transmission scenario, since basic isolation and social distancing can cut transmission. This could be much more difficult in a mirror bacteria scenario where pathogens could persist in soil, water, or dust. In these scenarios, maintaining distance from potentially infected people might be insufficient for survival since the environment itself could become contaminated and lethal—it would be like every person in the world becoming severely immunocompromised.

We know it’s possible to build infrastructure to keep immunocompromised people safe. For example, hospital cleanrooms achieve this level of protection through aggressive air filtration; David Vetter (a child with severe immunodeficiencies) was kept alive for 12 years using a sterile clean room and a plastic containment device (‘bubble’). Analogous protective infrastructure also exists in military contexts: most modern tanks and armored vehicles have intense filtration systems designed to protect against biological or chemical weapons, and modern militaries typically retain units with ‘collective protection’ equipment specialized for operating in environments with widespread biological or chemical contamination.

The problem lies in scaling this infrastructure to protect entire populations. Some countries, most notably Finland and Switzerland, invest so heavily in civil defense that all of their citizens already have access to bomb shelters that are also built to withstand incoming biological and chemical threats. While these shelters could potentially protect millions of people from a mirror bacteria-type scenario (at least as long as food reserves last), this would still only cover ~0.2% of the world’s population.

Fortunately, there are a number of promising paths forward that could potentially turn regular homes and office buildings into pathogen-resistant spaces with high levels of protection. Ultraviolet light (especially far-UVC) is one commonly cited example of a technology that could be widely adopted and wouldn’t require substantial retrofitting of existing infrastructure. Basic air filtration can also remove particles from the air.^[15]

It is also worth exploring whether large fractions of the world’s population could obtain high levels of protection on very short notice (e.g. within two weeks) by taking emergency measures. For example, triethylene glycol is a vapor commonly used in natural gas processing; the US already produces enough of it each year to continuously cover all US industrial floorspace and a fraction of residential floorspace as well.^[16] It seems effective at killing pathogens when deployed in the air or on surfaces, and also seems extremely safe for humans to breathe. More studies should be done to explore the efficacy and safety of this option (as well as others, e.g. propylene glycol), and emergency plans could be made to redirect and ramp up production in an emergency.

Perhaps most intriguing is the speculative possibility of ‘DIY biohardening’—where individuals or small groups use household materials to locally create cleanroom-like environments in an emergency. An initial investigation of the options and materials required has made this seem less hopelessly far-fetched than one might initially expect. For example, regular household insulation could potentially be converted into improvised HEPA filters, many households have furnace fans that are powerful enough to push air through such a HEPA filter and create positive pressure, and surfaces can be sterilized with hypochlorous acid—which can be made at home using salt water and an electric current.^[17] Stress-testing these interventions to validate them (or rule them out) would be an obvious next step.

Note that in a mirror bacteria scenario, biohardening is not particularly helpful without widespread access to PPE. Even if most people can safely shelter in place, a substantial fraction would still need to go outside to get food, power, and clean water, as well as to build more comprehensive medical countermeasures and other systems that could allow humanity to flourish in the longer run. Staying inside is not a viable long-term strategy.

Tentative victory condition: A guidebook and/or LLM assistant is developed to the point where a randomly selected group of people could successfully retrofit a building within two weeks with existing or stockpiled materials to achieve 3-5 logs of protection against environmental pathogens. If this is possible, it ought to be achievable by the end of 2027 with a concerted effort and perhaps $5-10M/year.

Detection

The previous pillars assume that society is taking a threat seriously (e.g. PPE only works if people are wearing it). One way this could fail is if a pandemic was spreading without showing many symptoms, with lethal symptoms emerging after it was too late. HIV is one example of a pathogen with a long latent period—severe symptoms typically don’t emerge until 8-10 years after an initial infection, and the virus was spreading for many decades before it was discovered in the 1980s. In a worst-case scenario, one could imagine something analogous to HIV that was airborne, where most of the world becomes infected before we realize how dangerous the situation is.

The third pillar is therefore ensuring that all pathogens are detected before they can infect a large fraction of the world population. To do this well against arbitrary biological threats, detection needs to be pathogen-agnostic and conducted frequently even if there is no apparent emergency. The current state of the art is metagenomic sequencing, which could in theory pick up any sufficiently prevalent pathogen with a DNA (or RNA) sequence; early efforts are already underway to pilot this technology against high-latency pathogens. Other technologies beyond metagenomic sequencing could also supplement this (e.g. protein sequencing, mass spectrometry, electron microscopy).

In the long run, detection should be biased in favor of the defender, so long as there are dedicated resources focused on the problem. If AI tools become advanced enough to re-design pathogens to evade screening, then they should also be equally capable of conducting the screening in the first place—in the same way that it is generally easier to verify a result than to generate it, it ought to be fundamentally easier for AI tools to verify that something is dangerous than to generate the dangerous thing in the first place.

Tentative victory condition: Any novel pathogen that could be created in the near future, no matter how stealthy or fast-spreading, is widely recognized at least a month before most of the world’s population would have become infected. Assuming a fast doubling time of 3 days and a very simplistic doubling model, this would mean detecting at ~0.1% cumulative prevalence. Initial programs could probably achieve a robust version of this by the end of 2027 with a concerted effort and less than ~$100M/year.

Expression of interest and acknowledgements

Reducing catastrophic biological risks is one of the largest challenges facing humanity this century, and the three pillars described above are one path forward. Turning this plan into reality will require more talented and ambitious people. If you are one of those people, reach out to us here.

This post is based heavily on research done by Adin Richards and Damon Binder. Many similar ideas previously outlined by Carl Shulman here.