# Radical Numerics Launches AI Tool to Detect Engineered Bioweapons The startup's new model identifies potential pathogens using portable, offline gene sequencing technology. - **Source:** [Wired](https://www.wired.com) - **Published:** 2026-11-11 - **Topic:** [technology](https://brifing.si/c/technology) - **Original article:** - **On Brifing:** https://brifing.si/s/03e16981-4e24-44d7-8b86-df664d6e18c9 ## New AI Biodefense Tool Radical Numerics has introduced Omnii Defense, an edge AI model designed to identify natural and AI-engineered pathogens. The tool operates on laptops connected to USB gene sequencers and functions without internet access. This capability allows researchers in remote field locations to monitor biological threats despite having limited or unreliable network connectivity. ## Foundational Research Origins The company was established in 2025 by former Stanford University researchers who previously developed Evo and Evo-2. These open-source DNA foundation models served as the technical basis for their current work. The team aims to address a perceived gap in the biosecurity sector by applying advanced machine learning to defensive pathogen surveillance. ## Genome Language Modeling Omnii utilizes a genome language model trained on diverse biological data, including RNA, proteins, and epigenetic signals. Unlike traditional tools that rely on exact sequence matches, this AI identifies patterns associated with biological functions. This approach enables the detection of threats even when genetic sequences have been modified by malicious actors. ## Threat Scoring Mechanism The model 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. By analyzing these complex variables, the system provides a quantitative assessment of risk that exceeds the capabilities of standard, library-based comparison tools. ## Experimental Performance Results In tests involving the Bundibugyo virus, Omnii successfully identified pathogenic sequences even when the virus was excluded from its training data. The model flagged 33 percent of sequences as pathogenic when only the species was removed. When the entire genus was omitted, it still detected 22 percent of the viral sequences as risky. ## Comparison With Legacy Tools Traditional biodefense tools like the federally funded Felix rely on comparing sequences against known organism libraries. These systems struggle when pathogens are novel or heavily modified. Radical Numerics claims its AI provides a step change by predicting potential risks rather than simply searching for cataloged matches that already exist in databases. ## Controversy and Transparency The company recently faced internal scrutiny after omitting a partnership with Tabula Bioscience during a public announcement. This lack of disclosure led Harvard geneticist George Church to resign from his advisory role at Radical Numerics. The incident highlights the complex collaborative landscape and transparency challenges within the emerging AI-driven biotech startup sector. ## Future Deployment Strategy Access to Omnii Defense is currently restricted to vetted external testers. Radical Numerics is actively collaborating with national laboratories and government agencies to validate the model's performance. While experts acknowledge the utility of such tools, they emphasize that detecting a signal remains distinct from confirming a pathogen's real-world consequence or origin. ## Sources covering this story - https://www.wired.com/story/startup-fights-ai-bioweapons-with-ai/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._