Cybersecurity startup Corma reaches unicorn status with $125 million Series A
TechCrunch|AIDEN Editorial|Nov 17, 2026|Updated Nov 22, 2026
Cybersecurity artificial intelligence startup Corma has raised $125 million in Series A funding led by NEA, valuing the firm at $1 billion just three months after emerging from stealth. The round reflects an urgent enterprise push for specialized defensive foundation models as general-purpose models demonstrate sharp imbalances between offensive capabilities and defensive detection.
Cybersecurity startup Corma has achieved a $1 billion valuation after securing $125 million in Series A funding led by venture firm NEA, reaching unicorn status roughly three months after emerging from stealth mode and unveiling its product. Additional investors in the round include Sequoia Capital, Khosla Ventures, and Coatue, which brings the company total funding to $185 million following an initial $60 million seed round raised in August 2026. The rapid influx of capital comes as enterprise demand surged, prompting Corma to expand its server infrastructure capacity by 24 times to handle requests from Fortune 100 and Fortune 500 organizations. Corma co-founder and chief executive officer Alon Pluda noted that global enterprise interest exceeded expectations, making infrastructure expansion an immediate priority.
The swift rise in defensive AI valuation aligns with increasing regulatory focus, such as the National Institute of Standards and Technology Artificial Intelligence Risk Management Framework, which encourages proactive threat management across digital networks. The shift follows an October 2026 security crisis affecting major open source software including PHP and Node.js, highlighting systemic defensive gaps. While legacy security vendors often integrated general commercial chatbots into administrative dashboards, those tools frequently struggled with high false positive rates and tracking complex threat signatures. Pluda observed that cybersecurity has evolved into an intelligence-against-intelligence contest, exposing the inadequacy of basic generative assistants in mitigating advanced intrusions.
Underlying this market demand is a significant capability gap revealed in testing conducted by Corma across simulated corporate networks. In experiments involving leading large language models, the models achieved an 85 percent success rate in breaching defenses and installing malicious files, but successfully detected those same implanted files from a defender perspective only 12 percent of the time, revealing a 73 percentage point detection deficit. Over the next six months, enterprise security leaders, chief information security officers, and cloud service providers are expected to reassess deployments of standard commercial model wrappers. Instead, organizational budgets are likely to pivot toward specialized foundation models and autonomous defensive agents capable of continuously parsing complex logs and telemetry in real time.
— AIDEN Editorial Team
What this means for the market
The rapid ascent of dedicated defensive model providers underscores a broader structural shift in enterprise software procurement, as general-purpose conversational models prove insufficient for critical cyber infrastructure. As frontier developer competition fuels sophisticated offensive capabilities, organizations and cloud providers must transition from reactive generative assistants to specialized autonomous defensive architectures. This dynamic is poised to accelerate specialized foundation model training across high-stakes industries, driving capital toward domain-specific inference workflows.
How this issue is unfolding
As frontier AI research laboratories focused heavily on coding and inference benchmarks, the offensive vulnerability-hunting capabilities of models advanced exponentially, while defensive capabilities for monitoring and addressing complex firewall logs and endpoint telemetry over extended periods lagged behind. Following a widespread vulnerability crisis in October 2026 that impacted open source software such as PHP and Node.js, enterprise and government awareness of this defensive gap heightened significantly. Traditional security vendors largely attempted to address the challenge by layering commercial chatbots over management consoles, resulting in high false alarm rates and clear limitations in analyzing long-term threat signals. By pre-training foundation models and autonomous agents tailored specifically to defensive workflows rather than wrapping general-purpose systems, Corma captures enterprise compute demand and illustrates how cybersecurity AI is pivoting from passive assistants to autonomous defense infrastructure.