Corma Raises $60M to Build Defensive AI

The big picture: AI models are increasingly effective at attacking computer systems, creating a widening gap in defensive capabilities. Corma, a cybersecurity AI startup, has raised $60 million in seed funding to address this, betting on a foundation model purpose-built for defensive cybersecurity rather than general-purpose AI.

Why it matters:

• Asymmetric Threat: Corma’s simulations show AI attackers succeeding 88% of the time, while AI defenders detected only 12% of threats, highlighting a critical imbalance.
• Operational Speed: AI-powered attacks operate at a speed and sophistication that human teams, traditional tools, or general-purpose AI cannot match, demanding an equally autonomous defense.
• Data Complexity: Defensive cybersecurity requires sifting through vast quantities of disparate logs, network traffic, and alerts, connecting signals over extended periods, a task poorly suited for general AI models.

How it works:

• Specialized Foundation Model: Corma is developing a foundation model specifically trained for the complexities of real-world defensive cybersecurity.
• AI Agent Workforce: This model powers AI agents designed to function as digital members of a security team, working across various defensive security functions.
• Early Deployment Impact: The company claims early deployments in Fortune 100 and Fortune 500 organizations have cut threat response times by over 94% and increased security coverage by 15 times.

The catch: Corma’s claims of drastic improvements in threat response and coverage are based on internal data and require independent validation as the company scales. As a six-week-old startup, it faces significant execution risk in a crowded market where many established cybersecurity vendors are also integrating AI. The ambitious bet that dedicated AI agents are required to fight other AI agents, rather than enhanced existing solutions, remains to be proven at a broad enterprise level.