The first permanent quarantine facility in the world was a brick plague hospital, or lazaretto, built on an island in the Venetian lagoon in 1425 to protect the city from the Black Death. Since then, the architecture of containment has failed repeatedly. Often, the escapes have been intentional, if not necessarily malicious. In the 1780s, a guard at the lazaretto in Split, Croatia, smuggled home a beautiful white scarf as a gift for his wife, inadvertently releasing bubonic-plague-infected fleas that killed one in 10 of the city’s residents. In the 1830s, a bored Boston ship’s pilot took advantage of the temporarily frozen harbor to walk from Quarantine Island to shore, triggering a cholera panic. Other times, pathogens have simply hitchhiked across boundaries with the help of unwitting human carriers; wheat stripe rust, a fungal disease that reduces harvests by up to 40 percent, is thought to have been brought to Australia on an international traveler’s trouser cuffs.

“Everyone will tell you, it always comes down to the people,” Trewyn said. They have been the Achilles’ heel of lazarettos throughout history. Of course, the NBAF is carefully engineered to lower the risk of human error. On the lab floor, people, animals, and stuff can move in only one direction, from clean to contaminated, “cold” to “hot.” Everything—including the animals, in carcass form—exits through a fumigation vestibule, a chemical dunk tank, or an autoclave, with the sole exception of people, who have to take two chemical showers and one regular shower before they can leave. (At Plum Island, researchers often complained that the showers had only a curtain separating them from the adjacent corridor. At the NBAF, decon happens in a personal air lock. “Times have changed,” Cole said.)

The NBAF will implement continuous training, detailed record-keeping requirements, and a buddy system, so that staff can inspect one another for trace contamination after showering out. Everyone who works there will have to pass background screening and security checks. The building has concentric rings of facial-recognition and PIN-code checkpoints. The precautions even extend into researchers’ private lives: They will not be allowed to keep chickens, on the off chance they bring home a pathogen that jumps species. Even so, the National Research Council has complained that Homeland Security’s risk assessment—a 0.1 chance of a pathogen escaping—was “based on overly optimistic and unsupported estimates of human error rates.” (The assessment didn’t even attempt to quantify the likelihood of malicious or deliberate acts.)

Certainly, Plum Island has had a handful of documented close calls, as have other such facilities around the world. But Trewyn believes that the risk of accidental pathogen escape, in all its uncertainty, is worth taking. These diseases will arrive in the United States anyway, he says, and cause equally incalculable damage.

Trewyn pointed to the very different courses taken by two outbreaks of the same disease in the UK—one caused by a lab leak, the other by an inadvertent introduction. In 2007, foot-and-mouth disease virus slipped out of the Pirbright Institute, an animal—disease research facility, into the Surrey countryside, with the help of heavy rains and poorly maintained pipework. It was quickly caught and contained. Within hours of the first case, the government halted all movement of livestock in the entire country; within two months, the virus was mopped up, having infected just eight farms. The system worked, Trewyn concluded, especially when compared with a very different incident six years earlier.