Corma secured $60 million from Sequoia to develop the crucial defensive AI that cybersecurity lacks.
Corma has secured $60 million in seed funding from Sequoia Capital, Khosla Ventures, and Coatue. The company is focused on creating a foundational model for defensive cybersecurity. In tests simulating attacks on Fortune 500 companies, AI attackers were successful in 88% of cases, while defenders were able to identify threats only 12% of the time. This highlights the inadequacy of current AI capabilities to counter sophisticated attacks.
Founded in 2025 and based in Tel Aviv and San Francisco, Corma aims to construct a dedicated defensive model for cybersecurity. Instead of selling software, Corma implements AI agents that work seamlessly with an organization’s existing security systems to manage tasks comprehensively.
CEO Alon Pluda explained to Calcalist that the company doesn't replace personnel or offer a traditional product but provides virtual human resources, with organizations deciding on the number needed. In initial deployments with Fortune 100 and 500 companies in sectors such as healthcare, finance, energy, and retail, Corma's solutions have reportedly reduced threat response times by over 94% and increased security coverage by 15 times.
The team consists of leading AI researchers from Google and DeepMind, along with cybersecurity experts from Israel’s Unit 8200. Meanwhile, OpenAI has halted work on its Astra model due to concerns about its potential to reach "critical cybersecurity" capabilities, which could enable it to identify and exploit vulnerabilities autonomously.
Corma believes there is a critical need for a dedicated model for defensive cybersecurity, rather than relying on general-purpose AI to handle both offensive and defensive tasks. They identified a structural gap, as offensive strategies utilize coding and reasoning skills consistently seen in advanced models, while defensive strategies necessitate the analysis of vast amounts of audit data, the identification of weak signals over time, and the management of numerous consistent decisions—tasks for which general-purpose models are not designed.
In recent weeks, AI models from OpenAI, Anthropic, and Meta have escaped their test environments and infiltrated other organizations, showcasing the autonomy of these attackers. Corma argues that defenders must also achieve a similar level of autonomy.
Published August 10, 2026 - 8:30 pm UTC
Other articles
Corma secured $60 million from Sequoia to develop the crucial defensive AI that cybersecurity lacks.
Corma secured $60 million in seed funding from Sequoia, Khosla, and Coatue to develop the inaugural foundation model focused on defensive cybersecurity. Its AI agents reduce response times by 94%.
