Radical Numerics, a San Francisco-based biotech startup, announced the release of Omnii Defense on Wednesday. The new tool is designed to identify natural and AI-engineered pathogens using only a laptop and a USB gene sequencer. The system operates without an internet connection, allowing researchers in remote areas with limited connectivity to analyze biological samples in real time.
The company aims to address a growing concern among frontier artificial intelligence labs that their technology could be misused to create dangerous biological weapons. While many scientists consider the threat of AI-made bioweapons to be low probability, Radical Numerics believes proactive defense measures are necessary. The tool assigns a threat score between zero and one to genetic sequences, combining the likelihood of unnatural origin with functional predictions.
Offline Analysis for Field Researchers
Omnii Defense runs on the company’s broader genome language model, Omnii. This model processes various types of biological information, including RNA, proteins, and epigenetic signals, rather than relying solely on DNA data. The offline capability is a key feature for public health officials and researchers working in the field. It allows for immediate analysis without the latency or security risks associated with cloud-based processing.
Eric Nguyen, cofounder and chief executive officer of Radical Numerics, described the current landscape as an emerging arms race in biodefense. He compared the situation to cybersecurity, where threat actors and defenders continuously adapt to gain an advantage. Nguyen stated that the biosecurity space has historically lagged in defensive capabilities compared to offensive potential.
The startup was founded in 2025 by Nguyen and his cofounders, who previously worked on open-source DNA foundation models called Evo and Evo-2 during their doctoral research at Stanford University. The company recently gained attention for using Omnii to create custom bacteriophages to treat a drug-resistant E. coli infection. This treatment received special permission from the Food and Drug Administration.
Performance Against Known Pathogens
Traditional biodefense tools, such as the federally funded Finding Engineering-Linked Indicators (Felix), compare genetic sequences against libraries of known organisms. These systems search for cataloged matches, which is useful for identifying tweaked versions of existing pathogens. However, they do not make predictive assessments about potential threats. Omnii Defense uses AI to learn patterns associated with biological function, allowing it to recognize similarities even when genetic sequences have changed.
In internal testing, the company removed the Bundibugyo virus from its training data. This virus is currently causing an Ebola outbreak in East Africa. When the model was tested with species-level removal, it identified 33 percent of the Bundibugyo sequences as potentially pathogenic. Even when the entire genus was removed from the training set, the model flagged 22 percent of the sequences. Non-AI tools provided no similar risk indications in these scenarios.
Christine Parthemore, Radical Numerics’ head of biodefense, described the model as a step change in detection capabilities. She noted that the AI can operate with a level of historical ignorance, similar to conditions in 1976 when Ebola was first discovered, yet still identify threats.
Access to Omnii Defense is currently restricted to vetted external testers. Radical Numerics is collaborating with national laboratories, government agencies, and independent evaluators to assess the model’s performance. Bronwyn MacInnis, director of pathogen surveillance at the Broad Institute of MIT and Harvard, emphasized the need for vigilance regarding AI-enabled biological risks. She cautioned that detecting a signal does not automatically confirm real-world consequences or intentional engineering.
The launch follows recent internal changes at Radical Numerics. Harvard geneticist George Church resigned from his advisory role last week. His departure came after the company announced its bacteriophage treatment without disclosing a partnership with Tabula Bioscience, another AI biotech startup involved in the research.
Source: Wired
