Corma emerged from stealth with a $60 million seed on August 10, 2026, led by Sequoia with Khosla Ventures and Coatue. The Tel Aviv / San Francisco lab is not selling another SIEM skin — it is training a defensive cybersecurity foundation model because general-purpose AI already helps attackers more than defenders.

Key facts

Who uses the product — and for what job

Users: CISOs and security operations leaders at large enterprises — especially regulated and critical-infrastructure environments.

Job: put an AI-native defensive workforce on logs, events, and network flows that general LLMs handle poorly, while attackers use those same models for exploit research and agentic attack chains.

Onboarding metaphor in the company’s materials: hire Corma like a new team member, not install another alert dashboard. That framing matters for buyers tired of tool sprawl.

Why now

• Frontier models got dramatically better at code reasoning — which maps cleanly to vulnerability research and exploit development.
• Anthropic’s Mythos-style disclosures made boards treat autonomous attack chains as a near-term risk, not sci-fi.
• Company sims claim an 88% vs 12% offense/defense success asymmetry when the same model class attacks then defends — a quotable wedge even if you treat the number as proprietary research.
• Offense-side cyber AI (continuous pentest, red-team agents) already raised mega-rounds; defense foundation models were the open lane.

Why Sequoia / Khosla / Coatue — portfolio fit

Likely founder rationale: raise a seed large enough to pre-train a domain model from a syndicate that already underwrites frontier AI compute and will not force a premature SOC-UI pivot.

Competitive map

Corma’s claim is category creation: defensive foundation model, not “ChatGPT for your SOC.”

When not to model this deal

• Wrong if you treat seed ARR claims as audited — early deployments are company-reported.
• Wrong if you assume general-purpose Claude/GPT fine-tunes close the gap without domain pre-training.
• Wrong if you conflate Corma with offensive AI pentest vendors — opposite buyer job.

Practical takeaway

• Founders (security AI): Prove defense on messy enterprise telemetry, not CTF toys; the asymmetry story is the memo.
• Investors: Underwrite data rights + model evals as carefully as growth — false negatives in critical infra are existential.
• Operators: Ask what Corma autonomously remediates vs escalates, and how it sits beside your existing SIEM/XDR stack.

Sources

1. ACCESS Newswire / FinancialContent (Aug 10, 2026): https://www.financialcontent.com/article/accwirecq-2026-8-10-corma-the-first-frontier-defensive-cybersecurity-ai-lab-raises-60m-as-ai-supercharges-attackers
2. Tech Startups Aug 10 roundup: https://techstartups.com/2026/08/10/venture-capital-startup-funding-roundup-august-10-2026-accel-coatue-management-general-atlantic-khosla-ventures-sequoia-capital-more/
3. Related: /startup/corma · /fund/sequoia · /fund/khosla-ventures · /fund/coatue