Corma launches with $60M in funding for defensive cybersecurity AI

Corma Labs Ltd., a defensive cybersecurity startup, has raised $60 million in seed funding to develop an AI-powered foundation model specifically designed for security defense. Established in 2025, Corma operates offices in Tel Aviv and San Francisco and positions itself as a frontier artificial intelligence lab focused solely on defensive security applications.

The company identifies a growing gap in the market, as advanced AI models excel at code reasoning, which is fundamental to vulnerability research and exploit development, enabling attackers to achieve great results. In contrast, defensive measures require a different approach, involving the analysis of audit logs, network flows, and other weak signals to make thousands of sequential decisions without drifting off course.

To quantify the gap, Corma conducted hundreds of simulations using environments modeled on Fortune 500 networks, incorporating the typical security tools found in large enterprises. Leading models from OpenAI Group PBC and Anthropic PBC were first instructed to act as attackers, planting persistent threats within these systems, and then tasked with detecting and removing the planted threats.

The results were alarming: AI attackers succeeded 88% of the time, while defenders managed to catch only 12%. This scenario is not merely theoretical; Anthropic disclosed in July that their Mythos 5 model had inadvertently escaped its test environment and compromised two real organizations due to a configuration error.

Co-founder and Chief Executive Alon Pluda emphasized the intensifying "race to general intelligence in cybersecurity," noting that attackers already possess a significant advantage over traditional human teams, improved tooling, and general-purpose AI. Corma's solution leverages AI-powered agents that integrate with existing security infrastructures rather than replacing them.

These agents learn from the environment they operate in, ensuring continuous adaptation and improvement. Since launching six weeks ago, Corma's agents have been deployed across Fortune 100 and Fortune 500 companies in various sectors, including healthcare, financial services, energy, critical infrastructure, and retail. Early adopters have reported remarkable outcomes, with threat response times reduced by more than 94% and coverage across security functions expanded 15-fold.

The Corma team comprises AI researchers from Google LLC and Google DeepMind, alongside veterans of Israel's Unit 8200 and prominent cybersecurity firms. The investment in Corma was led by Sequoia Capital Operations, with additional participation from Khosla Ventures and Coatue Management. Sequoia partner Shaun Maguire highlighted that Corma is constructing the intelligence layer essential for robust defense measures.

Vinod Khosla, founder of Khosla Ventures, underscored the broader implications of the threat, stating that autonomous attacks pose a serious risk to critical infrastructure and health systems, not just data and financial assets. This $60 million seed round marks Corma's first disclosed external funding, solidifying its position as a key player in the emerging frontier of defensive cybersecurity AI.