DATABASES Signal 450
Corma raises $60M seed after tests show frontier models detect only 19% of attacks
↗
AI security startup Corma raised $60 million in seed funding and published research showing four frontier models implanted persistent backdoors 85% of the time but detected only 19% of attacks across 241 scored engagements.
The asymmetry between offensive and defensive AI capabilities means organizations relying on general-purpose frontier models for defensive security tasks are significantly underprotected. As models improve at offensive security exponentially faster than defensive tasks, engineers need specialized tooling rather than repurposed general models to handle logs, events, and configurations effectively.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Corma tested Claude Opus 4.8, GPT-5.5, Grok 4.3, and DeepSeek V4 across 241 scored engagements, finding models implanted backdoors 85% of the time but detected only 19% of attacks.
The startup raised $60 million in seed funding led by Sequoia Capital, alongside Khosla Ventures and Coatue, and is working with Fortune 100 companies.
Defensive security tasks involve structured machine data like logs, events, configurations, and audit trails that models 'appear to read less reliably' than prose or source code.
THE CLUSTER