Cyber‑security startup Corma announced a $60 million seed round to build a defensive AI foundation model. Investors include Sequoia Capital, Khosla Ventures and Coatue, positioning the firm at the forefront of AI‑driven protection.

Key Takeaways

  • Corma raised $60 million in seed funding
  • The AI model is purpose‑built for defensive cybersecurity
  • Automated agents integrate directly with existing security stacks

Corma emerged from stealth on Monday, unveiling a $60 million seed round aimed at creating a defensive cybersecurity AI foundation model. Headquartered in Tel Aviv and San Francisco, the company was founded in 2025 with Alon Pluda serving as CEO.

Sequoia Capital, Khosla Ventures and Coatue led the round, providing the capital needed to process massive security telemetry—system events, audit logs, and network traffic—through a specialized AI engine. The model correlates subtle anomaly signals over extended periods to evaluate complex threats.

This architecture powers autonomous agents that embed into an organization’s current security infrastructure, working alongside human analysts to handle a wide array of defensive tasks while continuously learning from their environment.

Why This Matters

BozokMedia analysis shows that AI‑driven attacks are outpacing traditional defenses, making dedicated AI‑powered defensive workforces essential for enterprises to stay ahead of sophisticated threats.

Cyber‑security expert Dr. Maya Patel noted, "AI‑based defensive models can neutralize multi‑stage intrusions faster than any human‑only team."

Corma tested OpenAI and Anthropic models, finding they could launch end‑to‑end attacks but failed to defend against the same techniques. The company’s mission is to give defenders the same speed, sophistication, and generalization that attackers already enjoy.

Did You Know?: In 2024, only 12% of large enterprises used AI‑driven defensive solutions; by 2026 that figure is projected to exceed 45%.

Frequently Asked Questions

What types of data does Corma’s foundation model analyze?
It ingests system events, audit logs, network traffic, and other security telemetry to spot anomalies.

Can the AI agents work with existing security tools?
Yes, the agents are designed to integrate seamlessly with current security stacks and collaborate with human staff.