Corma Raises $60M From Sequoia: GPT and Claude Win 88% of Attack Simulations Against Defenders

Startup Corma raised $60M in seed funding from venture firm Sequoia in August 2026 to develop defensive AI, a new layer of enterprise cybersecurity that, according to the company, barely exists on the market today.

What hundreds of attack and defense simulations showed

The Corma team ran hundreds of simulations built on profiles of large Fortune 500 corporations, with dozens of real security tools in each environment. Each simulation tested leading language models, including GPT (OpenAI) and Claude (Anthropic). First, the models were tasked with attacking the organization and planting persistent threats. Then they were asked to swap roles and find what they had planted themselves.

The result was telling: the attackers won 88% of the time.

• Seed round: $60M, investor: Sequoia
• GPT (OpenAI) and Claude (Anthropic) were tested in both attacker and defender roles, taking turns
• Environment: hundreds of simulations based on Fortune 500 profiles with dozens of security tools
• Outcome: attacking AI won 88% of simulations

The result does not point to weakness in any particular model. GPT and Claude handled the attack task very well. It points to a structural asymmetry: offensive AI in cybersecurity is advancing quickly, while defensive AI remains in its infancy.

Why defensive AI lags behind offensive AI

Most current enterprise cybersecurity products follow a "detect and respond" model. They analyze incoming traffic, look for signatures of known threats and raise an alert. A proactive defensive agent, one that explores the infrastructure on its own, forms hypotheses about attack vectors and closes vulnerabilities autonomously, is effectively absent from the market.

Corma positions itself in exactly this niche. By the company's logic, SOC (Security Operations Center) teams can no longer keep up with the volume and speed of threats when working manually, and the only scalable answer is an AI agent that acts like an experienced analyst, without fatigue and with full coverage.

According to

Corma, as published by The Next Web, attacking AI wins 88% of simulated scenarios, including cases where the model planted the threat itself and then had to find it.

Who competes with Corma

No single player dominates the defensive AI segment for enterprise security yet, but the pressure is growing. Microsoft is pushing Security Copilot on top of its cloud ecosystem, while CrowdStrike and Palo Alto Networks are building AI layers into their SIEM and SOAR platforms. Corma's advantage is a clean focus on proactive defense and $60M in resources at the start from one of Silicon Valley's most respected firms.

A seed round of this size from Sequoia is rare: sums like this are usually raised at Series A. It suggests the firm is valuing not only the product but also the size of the untapped market. Enterprise cybersecurity globally generates hundreds of billions of dollars a year.

What this means

Corma's simulations are more than a marketing case. They are also an argument for the whole industry: attacking AI has already become an industrial tool, while the defensive layer is lagging badly. If the company manages to close that gap, it will hold a position in one of the fastest-growing enterprise AI segments.