Scientists have warned that the accelerating capabilities of artificial intelligence (AI) and synthetic biology are amplifying the risk of bioterrorism, after researchers discovered a “striking vulnerability” in software designed to control access to genetic materials used to produce potentially deadly proteins.
The international team behind the research described the flaw as the first “zero day” of AI and biosecurity—a term borrowed from cyber-hacking that denotes an unknown blind spot in software. While software updates have been deployed to close the loophole, experts say the episode highlights how rapidly advancing technologies can outpace existing safeguards.
“AI-powered protein design is one of the most exciting frontiers of science [and] we’re already seeing advances in medicine and public health,” said Eric Horvitz, Microsoft’s chief scientific officer and senior author of the study published in Science. “Yet, like many powerful technologies, these same tools can also be misused.”
The research team tested biosecurity software used by companies that sell synthetic nucleic acids—materials that clients use to build DNA instructing the manufacture of proteins. While the screening software was effective at flagging naturally occurring hazardous proteins, it struggled to detect altered versions generated using AI, which could evade safeguards by disguising their structure.
Even after most companies applied software patches, approximately 3 percent of the protein variants most likely to retain hazardous functionality still bypassed the monitoring system. Researchers worked with organizations including the International Gene Synthesis Consortium and U.S. authorities to address the vulnerabilities.
The study underscores the urgent need for improved global governance of AI-enhanced protein synthesis. High-profile biologists have called for international agreements to prevent the creation of potentially lethal “mirror” microbes, should the technology to manufacture them become feasible.
“By introducing targeted improvements to existing software, the authors significantly enhance detection and flagging,” said Francesco Aprile, associate professor in biological chemistry at Imperial College London. “This work provides a practical, timely safeguard that strengthens current DNA synthesis screening, and establishes a solid foundation for continued optimisation.”
Experts warn that the pace of technical progress in AI-driven biology will make these risks more acute. Natalio Krasnogor, professor of computing science and synthetic biology at Newcastle University, said the barriers that currently prevent aspiring bioterrorists from producing harmful proteins—expertise, time, and cost—are likely to shrink rapidly.
“We need as a society to take this seriously now,” Krasnogor said, “before additional advances in AI make the validation and experimental production of viable synthetic toxins much easier and cheaper to deploy than it is today.”
The findings highlight the intersection of cutting-edge science and national security, prompting calls for stronger, coordinated measures to protect global public health in the age of AI.
