Radical Numerics launches offline AI to flag engineered pathogens
What's the deal? Radical Numerics on November 11 unveiled Omnii Defense, an edge AI model it says can identify natural and AI-engineered pathogens using a laptop and a USB gene sequencer. The software runs without an internet connection, enabling field biosecurity screening, WIRED reported. Access is limited to vetted external testers.
What's the endgame? Rather than matching cataloged genomes, the model analyses functional patterns in biological sequences to spot unfamiliar risks. It generates a threat score from zero to one, combining functional predictions with the likelihood that a sequence is synthetic.
Why now? Omnii Defense extends the company's October launch of Omnii-1, a multimodal biology model it says it used to design bacteriophages given to a patient with a drug-resistant E. coli infection. That clinical outcome has not been disclosed, and the work has not been peer reviewed.
What could go wrong? The model's headline result — flagging 22% of Bundibugyo virus sequences when the genus was excluded from training — is a company test, not proof of accuracy or real-world effectiveness. Broad Institute researcher Bronwyn MacInnis cautioned that detecting a signal does not establish a pathogen's real consequence or whether it was engineered.
Radical Numerics is working with national laboratories and government agencies to validate results. A threat score flags a sample for investigation; it does not confirm the sample is engineered or dangerous.
The signal: The tool could complement database-based systems such as the federally funded FELIX, which can struggle with unfamiliar sequences. With access restricted and independent evaluation still underway, this is a notable biosecurity product preview rather than a validated detection breakthrough.
Read more: WIRED
Image credit: Radical Numerics
