Blog
by Aaron Jacobson, Thomas Joshi and Lila TretikovNov 16, 2026

We are living in unprecedented times in cybersecurity. Hackers have rapidly embraced AI’s advanced capabilities yet defenders are still early in adoption. While the attacker-defender asymmetry has long existed, “Black October” shows that gap has become unsustainable and we need a fundamentally new approach to cyber defense. Driven by this conviction, today we are excited to announce our partnership with a team bringing superintelligence to defensive cybersecurity: Corma.
Today’s frontier models, whether open or closed, are trained on code and human language. Since a key attack path for offensive security is finding and exploiting vulnerabilities, general models became excellent attackers almost by accident. Yet finding and fixing these vulnerabilities in advance to prevent breaches is only one part of a defender's job. They must also discover, investigate and respond to ongoing attacks by digesting logs, events, configuration state, etc. — a data distribution that general models barely see in training.
This blind spot accentuates the asymmetry already structurally inherent in cybersecurity. Attackers now have AI that generalizes, scales, and never sleeps. Mean time to exploit has gone negative, so exploitation is often occurring before the public is aware or a patch is issued.1 While defenders can use the same models to mitigate vulnerabilities proactively, they face the compounding forces of a large backlog, lagging enterprise patch cycles and a rising volume of attacks (including those beyond code exploits, like credential theft). AI threats will inevitably slip through so defenders must in turn use AI for defensive cybersecurity at machine speed, yet this work relies on capabilities that frontier models were never built for.
Enter Corma, the first frontier AI lab for defensive cybersecurity. The team is building a superintelligent foundation model trained natively on defensive security data and deployed as an autonomous AI security workforce.
One of Corma’s most interesting breakthroughs is adversarial training for cybersecurity. Inspired by the approach behind superhuman game-playing systems, it uses large-scale reinforcement learning in realistic enterprise environments, against a clear and verifiable outcome. Instead of learning about security from text, the model learns by doing the work. The result is a frontier defensive model that outperforms general-purpose LLMs on quality, speed and token cost.
Corma has also innovated many other leaps forward in training their model: how to treat security data as its own modality rather than forcing it into the shape of text and code, how to enable models to attend to the faint signals buried in a vast amount of noise, how to keep models extremely consistent across the thousands of decisions a single security investigation can require, and much more. Together, these advances are what set Corma’s model apart.
Beyond the technology, Corma has thoughtfully productized their model for frictionless deployment. The Corma Workforce, powered by the Corma foundation model, deploys within minutes and onboards like a new employee, using the same security and IT tools as the customer’s security team. It works alongside security teams through Slack, Teams, and email, just like another member of the team. The workforce autonomously executes complex security tasks end-to-end, proactively reporting findings and requesting approvals when needed.
Corma is already delivering measurable results across some of the world’s largest enterprises. Just three months after emerging from stealth, Corma already has multiple production deployments across Fortune 500 and large enterprises across healthcare, financial services, energy, critical infrastructure, and retail. Corma’s AI security workforce completes most security tasks autonomously and successfully, executes end-to-end tasks orders of magnitude faster, and enables security teams to complete 10x more work at much lower cost per task.2 Security tasks that previously went unaddressed due to limited resources are now executed autonomously, from start to finish. Most importantly, Corma has already autonomously detected, investigated, and contained active cyberattacks in real enterprise environments. Customers are rapidly expanding Corma’s responsibilities across defensive cybersecurity, from threat hunting and incident response to detection engineering, digital forensics, and beyond.
Corma’s mission requires fluency in both AI and cybersecurity. Founder and CEO Alon Pluda served in an elite intelligence sub-unit within Israel's 8200. Alongside Alon is a one-of-a-kind interdisciplinary team of top-tier hackers from industry and researchers from DeepMind, Microsoft, and other frontier labs.
We are proud to lead Corma’s Series A round and to partner with such a world-class team. We believe Corma can do for defensive cybersecurity what coding models did for developers, tilting the balance of power in favor of defenders.

Corma is hiring researchers and engineers, you can learn more at their website.
Sources
Disclaimer
The information provided in this blog post is for educational and informational purposes only and is not intended to be investment advice, or recommendation, or as an offer to sell or a solicitation of an offer to buy an interest in any fund or investment vehicle managed by NEA or any other NEA entity. New Enterprise Associates (NEA) is a registered investment adviser with the Securities and Exchange Commission (SEC). However, nothing in this post should be interpreted to suggest that the SEC has endorsed or approved the contents of this post. NEA has no obligation to update, modify, or amend the contents of this post nor to notify readers in the event that any information, opinion, forecast or estimate changes or subsequently becomes inaccurate or outdated. In addition, certain information contained herein has been obtained from third-party sources and has not been independently verified by NEA. Any statements made by founders, investors, portfolio companies, or others in the post or on other third-party websites referencing this post are their own, and are not intended to be an endorsement of the investment advisory services offered by NEA.
NEA makes no assurance that investment results obtained historically can be obtained in the future, or that any investments managed by NEA will be profitable. To the extent the content in this post discusses hypotheticals, projections, or forecasts to illustrate a view, such views may not have been verified or adopted by NEA, nor has NEA tested the validity of the assumptions that underlie such opinions. Readers of the information contained herein should consult their own legal, tax, and financial advisers because the contents are not intended by NEA to be used as part of the investment decision making process related to any investment managed by NEA