Radical Numerics Launches Omnii Defense to Counter AI Bioweapons

Radical Numerics, a San Francisco-based startup founded in 2025, has introduced Omnii Defense, a new edge model designed to flag both natural and AI-engineered pathogens. The system runs on a laptop paired with a USB gene sequencer and can operate entirely offline, making it ideal for field researchers in areas with spotty connectivity.

Unlike traditional biodefense tools such as the federally funded Felix platform, which relies on comparing sequences against known organism libraries, Omnii Defense uses deep learning to predict pathogenicity and AI generation likelihood. It assigns a threat score between zero and one, combining the probability of unnaturalness with functional predictions, allowing researchers to prioritize high-risk sequences.

The model goes beyond cataloged matches by learning patterns in biological sequences, enabling it to recognize similarities even when genetic spelling has been altered. This represents a step change in capability, as Christine Parthemore, head of biodefense at Radical Numerics, describes. In one test, the system demonstrated the ability to distinguish between natural and AI-modified genetic sequences with high accuracy.

Radical Numerics' broader portfolio includes the Omnii genome language model, which processes RNA, proteins, and epigenetic signals beyond DNA. The company previously developed Evo and Evo-2, open-source DNA foundation models during doctoral research at Stanford. In October, Omnii was used to create custom bacteriophages that targeted a drug-resistant strain of E. coli, resulting in a successful treatment for a sick patient under FDA approval. The startup's latest release, Omnii Defense, builds on this foundation to provide rapid, autonomous threat assessment in the field.