Corma has raised a $125 million Series A round at a $1 billion valuation, led by NEA, the company said. That is a striking number for a startup that, according to the report, only introduced its product three months ago. Corma offers AI models and agents trained for cybersecurity defense, and it has raised a total of $185 million from backers including Sequoia Capital, Khosla Ventures, and Coatue. Investors clearly see a large market here. The question is whether the speed of that money matches the maturity of the product it is funding.
The report ties Corma's rise to a day of alarming security news. On October 21, a critical flaw was found in PHP, a popular web programming language, and another was found in Node.js, the open source JavaScript runtime. The open source LiteLLM proxy, used for routing requests to multiple LLM providers, also disclosed a security incident, and other incidents surfaced the same day. The report describes this as an onslaught on security that AI has enabled. Read that way, fear of attacks became the sales pitch, and chief information security officers scrambling for answers made the market feel urgent.
Urgency is a poor basis for judging quality, and the report offers little independent evidence about what Corma's models actually do better than rivals. The report notes that Corma is not alone. It names Kevin Mandia's startup Armadin, along with Anthropic, OpenAI, Microsoft, and what it calls too many other startups to count. Corma says it stands out by being a frontier AI research lab focused on protection. That is a self-description, and the report does not show results that would confirm it. A company can be well funded and still be unproven.
The report itself says the market is young and necessary, and I agree that defense against AI-accelerated attacks matters. But necessity does not justify skipping scrutiny. Buyers of security products should ask for measurable outcomes, independent testing, and clear accounts of how these agents handle sensitive systems, not just valuation headlines. Investors should be asked the same questions before they write large checks. A billion-dollar number tells us what backers are willing to pay. It does not tell us whether the protection works.