FundingNovember 17, 2026 · 3 min read

Corma raises $125M at a $1B valuation to train AI that defends networks

NEA led the Series A, three months after Corma's $60 million Sequoia-led seed. The Tel Aviv and San Francisco startup trains foundation models for defensive security work and says demand jumped after October's wave of breaches.

Two analysts seen from behind in a dark security operations center, facing a wall of monitors showing network maps, a world map and traffic graphs
Illustration: Venture Block

Corma, a Tel Aviv and San Francisco startup that trains AI models for cyber defense, has raised a $125 million Series A at a $1 billion valuation, the company said on Nov. 16. NEA led the round, with existing investors Sequoia Capital, Khosla Ventures and Coatue joining. Corma has now raised $185 million in total.

The round comes three months after Corma came out of stealth in August with a $60 million seed round led by Sequoia. Fortune reported at the time that the company did not disclose its seed valuation, so the size of the step-up is not public.

A model built for logs, not code

Corma's argument is that the AI race has favored attackers. Frontier labs are pouring resources into coding models, and the company says a model that gets better at writing software also gets better at finding the bugs in it. Defense, in Corma's telling, is a different job: sifting audit logs, security events and network flows, linking weak signals across systems over long periods, and making thousands of decisions in a row without drifting.

CEO Alon Pluda made the same point to Fortune in August, saying most defensive work "doesn't have anything to do with coding" and is more about "looking at logs, audits, [and] finding the needle in a haystack."

Sequoia's investment note describes how the model is trained. Corma runs large-scale reinforcement learning in simulated environments that copy real enterprise networks, including their tools, telemetry and noise. Sequoia compares cybersecurity to Go or chess: a two-player game with a clear reward signal, whether or not the defender was breached. The firm says the resulting model beats general-purpose foundation models on defensive tasks at much lower per-token inference cost. Those performance claims come from Corma and its investors and have not been independently benchmarked.

Corma deploys the model as an "AI workforce" of agents that work inside a company's existing security tools, with roles that Sequoia says span security operations, identity management, and cloud and network security. In August, Corma told Fortune its model had cut threat response times by 94% at organizations that adopted it.

Demand after Black October

Corma ties the raise to a run of security incidents in October dubbed "Black October" in its announcement. TechCrunch reported that on Oct. 21 a critical flaw was found in PHP, the widely used web programming language, another surfaced in the Node.js JavaScript runtime, and LiteLLM, an open source proxy that routes requests to multiple LLM providers, disclosed a security incident. Other incidents were disclosed the same day.

According to TechCrunch, interest in Corma's product jumped after those disclosures sent chief information security officers scrambling. The company says demand has grown across Fortune 100, Fortune 500 and Fortune Global 500 companies, and that it has expanded its server capacity 24 times to keep up. The announcement did not include revenue or customer counts.

"Since we came out of stealth, demand from the world's largest organizations has been overwhelming, and one of our hardest problems has been securing capacity fast enough to keep up with the demand," Pluda said in the announcement.

A crowded field

Corma is far from alone in pitching AI for defense. TechCrunch points to Kevin Mandia's new startup Armadin, as well as Anthropic, OpenAI and Microsoft. The big labs are also building security-specific models: OpenAI's GPT-6 Cyber found 81 Firefox flaws in testing, and Mistral pitched its open-weight Large 4 at cyber defenders. Corma is betting that a lab focused only on defense can stay ahead of general-purpose models in this one domain.

The company says its team includes researchers from Google DeepMind and security specialists from Israel's Unit 8200. The deal follows other large AI rounds this month, including LangChain's $300 million Series C on Nov. 13.

"Corma is building the intelligence layer that defensive security needs," said Aaron Jacobson, a partner at NEA. The company says it will use the money to scale its foundation models and the agent workforce built on them.

Newsletter

The Block Brief

The stories that matter in startups and venture, in your inbox. Free.

Check your inbox to confirm.

Something went wrong. Try again.

More in Funding