Corma raises $60M from Sequoia to be a frontier lab for defensive cybersecurity only
The Tel Aviv and San Francisco startup positions itself as the first frontier lab dedicated to defensive security AI. Khosla Ventures and Coatue joined the seed round.
Sixty million dollars is a large seed and a small frontier-lab budget, which locates the bet precisely: Corma is not training a foundation model, it is building the layer that turns one into a defender. The scarce input in security is not raw capability — it is context about a specific network, and the judgement to act on it without breaking production.
The timing next to OpenAI's Daybreak Red tier is not a coincidence. Once a general lab ships a model at 95 percent completion on advanced offensive-security tasks, the defensive side is no longer competing on capability access. It is competing on deployment, and a company whose only product is defence can make integration commitments a general lab will not.
The strategic question is whether "defence only" survives contact with revenue. The techniques are the same; the distinction is authorisation and intent, both of which live in customer contracts rather than in the model. Plenty of security companies have started with that line and found it moved.
AI Weekly — AI News Today, August 10 — Top AI Stories & Live Updates →