Research

Radical Numerics Launches AI Model to Detect AI-Generated Pathogens

A biosecurity startup unveils Omnii Defense, an edge model that identifies engineered biological threats without internet connectivity.

Last verified:

Omnii Defense: Detecting Engineered Threats at the Edge

Radical Numerics unveiled Omnii Defense on November 11, an edge-deployable AI model engineered to identify both naturally occurring and artificially constructed pathogens without requiring internet connectivity. According to Wired AI, the tool operates on standard laptop hardware paired with a USB gene sequencer, enabling field researchers and public health officials in regions with limited network access to screen genetic sequences in real time.

The model generates multiple risk assessments for each sequence analyzed, including predictions about pathogenicity, signs of AI engineering, and taxonomic classification. Omnii Defense then consolidates these signals into a single “threat score” ranging from 0 to 1, where higher scores indicate greater concern. This design distinguishes it from federally funded alternatives such as Finding Engineering-Linked Indicators (Felix), which rely on matching suspect sequences against catalogs of known organisms rather than identifying novel or modified variants through learned pattern recognition.

The Biosecurity Arms Race Framing

Eric Nguyen, Radical Numerics’ cofounder and CEO, has positioned the company’s work within a broader competition between threat actors and defenders—a dynamic already visible in cybersecurity. According to Wired AI, Nguyen argues that “the biosecurity space was far behind on the defense side,” motivating Radical Numerics to accelerate development of detection capabilities.

Radical Numerics’ leadership team includes researchers who authored Evo and Evo-2, open-source DNA foundation models developed during their Stanford doctoral work. The company was formally established in 2025 and is based in San Francisco. The organization has already gained attention for its primary product, Omnii, a broader genome language model trained on RNA, proteins, epigenetic signals, and DNA sequences. In October, Radical announced that Omnii had been used to design custom bacteriophages—viruses that target and kill specific bacteria—tailored to a drug-resistant strain of E. coli, which were subsequently administered to a patient under special FDA authorization.

Transparency and Trust Issues

The October bacteriophage announcement omitted disclosure that Tabula Bioscience, another AI biotech startup, had contributed significantly to the research. According to Wired AI, this lack of transparency prompted Harvard geneticist George Church—who served as an adviser to both companies—to resign from his role at Radical Numerics last week. Church’s departure signals internal disagreement over how such breakthroughs should be communicated to the public and the scientific community.

Why This Matters

The biosecurity community remains divided on the imminence of AI-enabled bioweapon development. Wired AI notes that “many scientists say AI-made bioweapons are a low-likelihood threat,” yet frontier AI labs have publicly acknowledged that their systems could plausibly be misused in this way. If Omnii Defense proves reliable in field conditions and under independent scrutiny, it could become part of a broader detection infrastructure for public health agencies and research institutions. However, the technology’s dual-use nature—the same pattern-recognition capabilities used to flag threats could inform offensive research—means its release into the field requires careful governance. The resignation of an advisor as prominent as George Church suggests that questions about disclosure, attribution, and responsible development remain unresolved within the biosecurity startup ecosystem.

Frequently Asked Questions

What is Omnii Defense and how does it work?

Omnii Defense is an edge-deployable AI model that analyzes genetic sequences and assigns a 'threat score' between 0 and 1, combining predictions about unnaturalness with functional analysis. It runs on a laptop with a USB gene sequencer and requires no internet connection.

Who founded Radical Numerics?

Eric Nguyen and cofounders, including researchers behind Evo and Evo-2 DNA foundation models, founded Radical Numerics in 2025. They conducted their doctoral research at Stanford University.

How does Omnii Defense differ from existing biodefense tools?

Unlike tools such as Felix, which match sequences against known organism libraries, Omnii Defense uses AI to recognize patterns and similarities between sequences even when they are novel or modified, enabling detection of engineered variants.

What controversy has surrounded Radical Numerics?

After the company announced bacteriophage treatment for drug-resistant E. coli in October without disclosing its partnership with Tabula Bioscience, Harvard geneticist George Church resigned from his advisory role at Radical.

#biosecurity #ai-safety #biotech #defense #foundation-models