Could AI Make Biological Weapons More Likely? Experts Weigh the Risk

Fully autonomous laboratories do not yet exist, so an AI cannot currently access a completely autonomous lab and build a virus on its own, said Olivia Sharfman, a biotechnology researcher at the Institute for Progress, a technology think tank. However, she warned that AI could potentially help someone hire another person to carry out the work.

Why Some Researchers Fear AI-Assisted Bioterrorism

Sharfman believes AI-assisted bioterrorism is an imminent threat and that some groups could be nihilistic enough to pursue it. She specifically cited “transhumanist AI successionists,” a fringe ideology whose supporters want to replace humans with bots.

Experts Disagree About the Value of Biological Weapons

Other experts are less concerned about AI bioweapons because they believe biological warfare is not particularly practical. Genetic biologist and computational biology professor François Berou told WIRED that the risks are “somewhat misunderstood,” adding that people tend to “overestimate the value of pathogens as weapons.”

Regardless of artificial general intelligence, biological weapons may not be an attractive option for would-be genocides, Bellu said. It is difficult to target specific groups while avoiding others, and mass-producing and distributing pathogens is logistically complex compared with detonating a bomb.

“If you want to kill someone, there are much, much better ways than creating and releasing viruses and bacteria,” Bellu said—whether the attacker is the most sophisticated artificial brain ever created or an ordinary human being.

How to Reduce the Risk of AI-Enabled Biothreats

As AI brings the threat of biological weapons into the spotlight, Sharfman said it is a good time to strengthen defenses against all biological threats, including existing pathogens such as H1N1. Potential measures include legislation to improve security protocols for DNA synthesis technology and upgrades to air purification systems in buildings.

Steph Guerra, director of AI and biotechnology at the RAND Corporation, said it is difficult to determine whether AI clearly increases the risk of developing biological weapons. “AI is proving to be very good at integrating information and motivating people to do good and bad things,” she said.

Even if the overall risk is low, Guerra said society can take steps to limit potential misuse. Governments could require companies that sell synthetic DNA and RNA to screen customers and provide instructions designed to prevent the misuse of genetic material. Many companies already use software to screen orders for potential “concerns,” but the practice is neither universal nor mandatory.

Most experts agree that AI models should include appropriate safeguards so malicious users cannot obtain actionable, dangerous information.

Why Biosecurity Requires a Layered Defense

Guerra also said the risk of AI-developed biothreats could be reduced by strengthening global disease-surveillance systems. Faster detection of new outbreaks and quicker access to pathogens for researchers could help limit the danger. Data-sharing protocols involving AI companies, gene-synthesis providers and governments could also help prevent abuse.

“For almost every biosecurity control, there will be ways around it,” Guerra said. “That’s why we need a multi-layered approach that creates friction along the entire path, from the bad actor’s idea to his intent to the release of the biological weapon.”

AI Could Also Accelerate Medical Advances

Unatmaz said there are more pressing concerns. He added that public discussion of AI’s potential for catastrophe can distract from the ways cutting-edge systems may assist humans with vaccine development and other medical advances.

“We really need to focus on the positive aspects of it,” he said.