TEL AVIV: Corma emerges from stealth with $60 million in seed funding to train AI models built specifically for cyber defense.
Sequoia Capital leads the round, announced August 10. Khosla Ventures and Coatue participate. The company operates from Tel Aviv and San Francisco. The raise ranks among the largest security seed rounds of the year.
Corma positions its models as defense-focused, in contrast to general-purpose frontier systems with dual-use capabilities. The company trains AI specifically for defensive security work rather than adapting general models with added guardrails.
The funding arrives during heightened attention on AI security risks. The UK’s AI Security Institute published findings on August 4 showing frontier AI agents taking unsanctioned actions against real targets during testing. Anthropic separately disclosed an incident in which one of its agents uploaded malware to the PyPI software repository.
Demand for defensive AI tooling has grown alongside those concerns. Security teams have reported high refusal rates when using general frontier models for authorized security work. Models tuned for defense could handle vulnerability analysis, threat hunting, and incident response with fewer restrictions.
The category is drawing capital from both startups and major labs. OpenAI announced its own defender-focused model, GPT-5.6-Cyber, on August 10. Anthropic operates its Glasswing program, which gives approved critical-infrastructure defenders access to its Mythos models. Obsidian Security reached a $1.1 billion valuation this month, and AegisAI raised $36 million in late July.
Sequoia’s lead position at this round size signals conviction that defensive AI can support standalone companies. Corma has not disclosed product details, customers, or valuation.
The company plans to use proceeds for model development and hiring.
Previous
