Risky business
There was commotion in the biosecurity world—that is, the community of people trying to prevent pandemics—when GPT-4 was released. Artificial intelligence intersects, or will intersect, with every realm of human activity. It could make pandemics a lot deadlier.
A report from NTI Bio, an American non-profit that develops policy-based solutions to pandemic risks, lays out how AI will threaten global biosecurity. Its large language models (LLMs) could democratise knowledge of how to assemble dangerous bacteria and viruses. Worse, it could plausibly be used to design and manufacture entirely new pathogens.
These are troubling prospects, and I do not envy Jaime Yassif her remit. Yassif, who co-authored the report I mentioned, oversees NTI Bio’s work on global health security. If there are such things as rock stars of pandemic prevention, she is one of them, appearing last year at a US House of Representatives hearing on America’s vulnerability to future pandemics. (Asked by Representative Debbie Dingell whether the US is indeed prepared, Yassif was blunt: “No.”)
I’m speaking to her shortly after her visit to Geneva for a meeting of the states that are signatories to the UN’s Biological Weapons Convention. She and her colleagues briefed delegates on the report into AI and biosecurity. “That was a very active discussion,” she says.
Since the outbreak of Covid19, Yassif has found it easier to be heard. Before the pandemic, she says, “It was hard to get traction.” Now, though, “There is much broader recognition among the public, and in policymaking circles, that a larger event like that could happen.”
Covid was devastating. But the next pandemic, she says, “could be as bad, or it could be orders of magnitude worse.”
Yassif saw this coming early. At the beginning of her career, she worked on nuclear security. But from the early 2000s, she had known that synthetic biology—the field in which scientists redesign existing organisms, or create new ones—was going to be consequential.
Her forecast was correct. As Yassif took increasingly senior biosecurity roles in government and nonprofits, it became easier and cheaper for scientists to assemble dangerous new pathogens in labs. And the number of labs handling the deadliest pathogens has multiplied, even though such labs have let those pathogens escape many times. Now, advanced AI is on the horizon.
How should we handle all these worries? “I would actually say I’m pretty analytical about
it,” says Yassif. “I don’t think I could work effectively if I was very emotional.” She refers to the human devastation caused by Covid and the bombing of Hiroshima in 1945. “You can’t lose sight of that. But when I’m in my professional mode, I’m thinking about it as a system.”
And it is systemic solutions for which she and NTI Bio advocate. “There is a tonne of work that needs to be done to develop new and innovative and practical approaches to meaningfully reduce risks. Interventions that have teeth, that actually change the game.”
One such intervention is the introduction of software to help DNA synthesis providers—companies that sell scientists the genetic material from which they make new organisms— cheaply and easily screen customers and their orders. (You’d hope that alarm bells would ring if a Mr Kim, from the Democratic People’s Republic of Korea, ordered smallpox DNA to Pyongyang.)
Another is the development of a diplomatic and scientific procedure by which experts, with UN backing, can investigate the source of new outbreaks. The Joint Assessment Mechanism, as this diplomatic apparatus is known, would resolve ambiguity over whether an outbreak is emerging naturally—or the result of a deliberate or accidental release. The prospect of such scrutiny, it is hoped, would prompt governments to behave responsibly.
Yassif praises the British government’s new biosecurity strategy. The UK, she says, is “showing leadership”. But there are plenty of ways in which things could go wrong. Let’s hope Yassif & Co can keep a lid on things.