The sample was only discovered when it started killing chickens.

In the spring of 1979, a lab worker in Sverdlovsk, USSR removed a clogged air filter in the ventilation system and didn’t replace it. His note to the supervisor was never transferred to the official logbook, so when the next shift rolled in, workers simply started production as usual. Now, in most labs, this would have been a minor mistake. But this lab was a biological weapons facility producing huge quantities of anthrax— which, if inhaled, can kill up to 90% of those it infects. This deadly anthrax powder floated out into the sky for hours, causing the largest documented outbreak of inhalation anthrax on record and resulting in at least 64 deaths. What happened at Sverdlovsk was a tragedy, and the Soviet bioweapons program was a violation of international law.

But these days, it’s not just state-sponsored bioweapons programs that keep biosecurity experts up at night. Nor is anthrax their largest concern. They’re worried about an even more dangerous kind of lab leak. Since the 1970s, researchers have been manipulating the DNA of microbes to give them abilities they didn’t have before. This is called “gain of function” work and it includes a huge body of scientific research. The majority of this work helps humanity with very little risk, for example, engineered viruses are used in vaccine production, gene therapy, and cancer treatments.

But within the gain of function realm lies an intensely debated sub-field where scientists engineer superbugs. Officially known as “enhanced potential pandemic pathogens” (ePPPs), these are typically variants of well-known viruses, such as Ebola or avian influenza that have been engineered to be, say, more transmissible or more deadly. The stakes of this kind of work are much higher: if even one unusually dangerous virus escaped a lab, it could cause a global pandemic.

Virologists developing ePPPs argue this research could help us prepare for future pandemics, allowing us to jump start treatments and potentially save lives. For example, in the early 2010s, several research teams created a deadly strain of bird flu with the novel ability to spread through the air between mammals. Advocates of the project argued that by creating this ePPP, we could learn crucial information about a worst-case-scenario virus under controlled conditions. But many critics argued that it’s unclear whether bird flu would ever evolve in the wild as it did in the lab. Consequently, they believed the knowledge gained by studying this dangerous virus wasn’t remotely worth the risk of creating it in the first place.

Both sides of this ongoing debate are trying to save lives; they just disagree on the best way to do it. However, everyone agrees that an ePPP lab leak could be catastrophic. Labs that work with dangerous pathogens are designed with numerous safety features to protect the scientists who work there, as well as the outside world, such as ventilation systems that decontaminate air and airtight “spacesuits” with dedicated oxygen. Sometimes buildings are even nested inside each other to prevent natural disasters from breaching the closed environment. But this technology is expensive to build and maintain.

And even when our tech doesn’t fail, there’s still room for the most common kind of mistake: human error. Many human errors are inconsequential: a researcher spills a sample, but quickly disinfects the otherwise well-controlled environment. Other incidents, however, are much more concerning. In 2009, a researcher accidentally stuck themselves with an Ebola-contaminated needle, endangering their life and the lives of those treating them. In 2014, six vials containing the virus that causes smallpox were found in an unsecured storage room where they’d been forgotten for decades. That same year, a CDC scientist unknowingly contaminated a sample of relatively harmless bird flu with a deadly lab-grown variant, and then shipped the contaminated sample to the USDA. The sample was only discovered when it started killing chickens. Despite the fact that these incidents did not result in major crises, the potential catastrophic outcomes of an ePPP leak has caused many researchers to advocate for ceasing this kind of research. However, if this is not possible, what can be done to lessen the risk? One suggestion is to reduce human error by analyzing past errors. Additionally, creating an international database of leaks, near-misses and fixes taken could help labs adjust their protocols to minimize human mistakes. Moreover, a strong and well-funded pandemic early warning system would help protect us from any disease outbreak, be it from a laboratory leak or a natural spillover. Implementing the global standards and databases required for these changes would be a difficult process, requiring unprecedented international collaboration and transparency. Nevertheless, it is essential to overcome these obstacles as pandemics do not differentiate between borders or politics.