Startup unveils AI tool to detect potential bioweapons from genetic data

Radical Numerics has introduced Omnii Defense, a new edge model capable of identifying both natural and artificially engineered pathogens using only a laptop and a USB gene sequencer, allowing offline use for researchers. Co-founder and CEO Eric Nguyen noted that the company views the biosecurity sector as lagging in defense, anticipating an arms race in biodefense akin to cybersecurity. The company, founded in 2025, previously released Omnii, a genome language model that successfully created custom bacteriophages to treat a drug-resistant E. coli strain with FDA permission, a project involving Tabula Bioscience. Omnii Defense assigns a threat score based on a sequence's likelihood of being unnatural or pathogenic. Christine Parthemore, head of biodefense at Radical Numerics, demonstrated its power by showing that even when removing the Bundibugyo virus from training data, the model still flagged 33 percent of related sequences as pathogenic, a capability non-AI tools lacked. While Bronwyn MacInnis, director of pathogen surveillance at the Broad Institute of MIT and Harvard, stressed the need for vigilance against AI-enabled threats, she cautioned that detection does not equate to determining real-world consequence. Access to the new tool is currently restricted to vetted external testers as Radical Numerics collaborates with national laboratories and government agencies.