Radical Numerics Launches AI Tool to Detect Engineered Bioweapons
Omnii Defense identifies potential biological threats using offline gene sequencing and predictive modeling.
1New AI Biosecurity Defense
Radical Numerics has introduced Omnii Defense, an edge model designed to identify natural and AI-engineered pathogens. The tool operates on a laptop connected to a USB gene sequencer, functioning without internet access. This capability allows researchers in remote field locations to screen biological samples for potential threats in real time.
2Predictive Threat Scoring
Omnii Defense assigns a threat score between zero and one to genetic sequences. This score integrates the likelihood of a sequence being unnatural with functional predictions regarding its pathogenicity. Unlike traditional tools that rely on exact library matches, this model identifies patterns associated with biological function to detect novel, modified threats.
3Experimental Performance Results
In tests involving the Bundibugyo virus, the model successfully flagged pathogenic sequences even when the virus was excluded from its training data. When the entire genus was removed, the model still identified 22 percent of viral sequences as pathogenic. Traditional, non-AI tools provided essentially no indication of risk in these specific scenarios.
4Foundational Research Origins
The startup was founded in 2025 by researchers who previously developed Evo and Evo-2, open-source DNA foundation models at Stanford University. CEO Eric Nguyen views the current biosecurity landscape as an emerging arms race, comparing it to the rapid evolution of cybersecurity where defenders must leverage AI to counter sophisticated threat actors.
5Clinical Application History
In October, the company utilized its Omnii model to create custom bacteriophages targeting a drug-resistant E. coli strain. With special permission from the Food and Drug Administration, this cocktail was used to treat a sick patient. This application demonstrated the model's ability to process biological data beyond DNA, including proteins and RNA.
6Internal Governance Challenges
The company recently faced internal friction regarding transparency. Radical Numerics failed to disclose a partnership with Tabula Bioscience during a public announcement, leading Harvard geneticist George Church to resign from his advisory role last week. This incident highlights the complexities of collaboration and disclosure within the competitive AI biotech startup ecosystem.
7Expert Perspectives on Utility
Bronwyn MacInnis of the Broad Institute notes that while using technology to compete with nefarious actors is sensible, detection remains distinct from confirmation. She cautions that identifying a signal is a significant leap from determining whether a pathogen has real-world consequences or if a specific threat was indeed intentionally engineered.
8Current Access and Deployment
Access to Omnii Defense is currently restricted to vetted external testers. Radical Numerics is collaborating with national laboratories, government agencies, and independent evaluators to rigorously assess the model's performance. These partnerships aim to validate the tool's effectiveness before broader deployment, ensuring it meets the safety standards required for sensitive biosecurity applications.
Original article: Science Latest — Radical Numerics Launches AI Tool to Detect Engineered Bioweapons