Corma

Corma is a foundation model lab focused on defensive cybersecurity, launching out of stealth with an AI workforce that plugs into an enterprise security team's existing stack and takes on the work that sits outside code. That work looks like sifting through audit logs, events, and network flows, correlating weak signals across long time horizons, and staying consistent across thousands of sequential decisions. Early customers span Fortune 100 and Fortune 500 organizations across healthcare, financial services, energy, critical infrastructure, and retail, with the first model deployed roughly six weeks before launch. The timing reflects a widening gap the company has tried to quantify. In internal simulations run against realistic Fortune 500 environments using models like OpenAI's GPT and Anthropic's Claude, AI attackers succeeded 88 percent of the time while AI defenders detected only 12 percent, meaning the same model that could carry out an end-to-end intrusion typically could not catch itself. Corma is pitching a purpose-built alternative it says outperforms general-purpose models on defensive tasks, runs faster and cheaper, and can be deployed fully on-prem and post-trained on a customer's own security data. The company, based in Tel Aviv and San Francisco and founded in 2025, is led by co-founder and CEO Alon Pluda, who previously served in Israeli military intelligence. It is announcing a $60M seed round led by Sequoia Capital with Khosla Ventures and Coatue also participating. The name comes from Tolkien's Elvish word for "ring," which Pluda frames as a tool built this time for the defenders.

Tags

Comments (13)