There is a growing gap in enterprise cybersecurity, and it is not about the tools you have. It is about who is using AI faster. Attackers are using it to probe systems, craft phishing campaigns, and launch multi-stage intrusions at machine speed. Most defenders are still relying on human analysts who can only work so fast.

A new startup called Corma wants to close that gap. On August 10, the company announced a $60 million seed round led by Sequoia Capital, with Khosla Ventures and Coatue joining. The thesis: if AI is supercharging attackers, the only answer is AI purpose-built for defense.

What Corma Is Building

Corma describes itself as the first frontier AI lab working exclusively on the defensive side of cybersecurity. The company is building a foundation model purpose-built for threat detection, response, and security operations — not a wrapper on top of a general-purpose model, but a model trained from the ground up on security data.

The team blends AI researchers from Google and DeepMind with cybersecurity specialists from Israel’s Unit 8200 intelligence corps and major security firms. The company was founded in 2025 and splits its operations between Tel Aviv and San Francisco, with about 20 people currently on staff.

Despite launching only six weeks before the funding announcement, Corma had already deployed its systems at Fortune 100 and Fortune 500 organizations across healthcare, financial services, energy, critical infrastructure, and retail.

The Numbers That Got Sequoia Interested

Early deployments produced results that are hard to ignore:

• Threat-response times reduced by more than 94%
• Security coverage expanded 15 times across different security functions
• Multi-stage attack campaigns uncovered that would have otherwise gone undetected

A 94% reduction in response time is not incremental improvement. When a breach is underway, the difference between minutes and seconds can be the difference between a manageable incident and a catastrophic one.

Why This Matters for Businesses Deploying AI

If your business is deploying AI agents — for customer service, operations, data analysis, or any other function — you are also expanding your attack surface. AI agents connect to internal systems, process sensitive data, and operate with degrees of autonomy that create new security risks. Every integration is a potential entry point.

The problem is compounded by the fact that attackers have access to the same AI tools as defenders, and they are not slowed down by compliance requirements or procurement cycles. They iterate faster.

Traditional security tools were not built for this environment. They were designed to detect known threats, not to reason over novel attack patterns in real time. That is exactly what Corma is trying to solve.

For enterprise and mid-market businesses, the practical implication is clear: the AI security threat is real and growing, and the response has to match it in kind. Having humans reviewing alerts on a dashboard is no longer enough when attackers are automating the entire kill chain.

The Bigger Picture

The Corma funding is part of a broader pattern. Venture capital investment in AI security companies has accelerated throughout 2026 as enterprises wake up to the gap between AI adoption and AI security readiness. Most organizations are racing to deploy AI agents, but fewer have thought seriously about what happens when those agents are targeted.

Corma’s focus on building a purpose-trained model rather than adapting a general AI is the right instinct. Cybersecurity is a domain where domain-specific training matters enormously. A model that has seen millions of attack patterns, understands attacker behavior, and can reason across a full network topology will outperform a generalist model given a security prompt.

What This Means for Business Leaders

If you are a business leader adopting AI — and the evidence says you either already are or will be soon — here is what the Corma story tells you:

AI security is not a nice-to-have. As you deploy more AI, you become a more attractive target for AI-powered attacks. Your security posture needs to evolve in parallel with your AI capabilities.

Specialist AI beats generalist AI in security. Just as Enterprise DNA builds domain-specific AI agents for business operations, the security world is figuring out that a model trained for a specific task performs better than a general one given instructions.

The talent gap is real. The reason Corma can cut threat-response times by 94% is not because human analysts were bad at their jobs. It is because there are not enough of them, and they cannot scale the way AI can. The same logic applies to any business function where demand for skilled work outpaces supply.

The $60 million Sequoia bet on Corma signals confidence that defensive AI is about to become a major category. For businesses deploying AI agents today, that is both a warning and an opportunity.