Radical Numerics unveils offline model for pathogen detection
The AMW Read
The offline model adds a pathogen-surveillance application, but limited access and pending independent evaluation constrain its demonstrated impact.
Radical Numerics unveils offline model for pathogen detection
Radical Numerics has introduced Omnii Defense, an AI model designed to flag natural and AI-engineered pathogens using a laptop and a USB gene sequencer. The model can run without an internet connection. It estimates whether a genetic sequence is pathogenic or artificially generated, identifies its closest taxonomic group, and combines predictions into a threat score. Access is currently limited to vetted external testers.
The launch extends the market for biological foundation models into field surveillance and biodefense. Radical Numerics says its approach can identify functional patterns in sequences rather than rely on exact matches to known organisms. In a company experiment that excluded Bundibugyo virus from training data, the model flagged 33% of its sequences as potentially pathogenic. When the company excluded the entire genus, it flagged 22%. Independent evaluation is underway, and those results do not establish the model's performance in field use.
For builders and investors, the key question is whether sequence-level signals can support reliable decisions outside controlled tests. A Broad Institute researcher cautioned that detecting a signal is a substantial step away from determining whether a pathogen has real consequences or was engineered. Radical Numerics says national laboratories, government agencies, and other independent evaluators are assessing the model's performance.