Skip to content
NewsFunding & M&A

Sequoia leads $60 million seed round into Corma, an AI lab building autonomous cyber defenders

· by Pondero Newsdesk

The short version

Corma raised $60 million in seed funding led by Sequoia Capital to build a foundation model for cyber defense. The company's simulations found AI attackers succeeded 88% of the time while AI defenders caught just 12%.

Sequoia leads $60 million seed round into Corma, an AI lab building autonomous cyber defenders

In simulations modeled on Fortune 500 networks, AI attackers succeeded 88% of the time. The AI defenders caught 12%. That gap is what Corma raised $60 million to close.

What

Corma Labs Ltd. announced its first outside funding on August 10, 2026: a $60 million seed round led by Sequoia Capital Operations, with Khosla Ventures and Coatue Management also participating, per SiliconAngle. The startup, founded in 2025, operates from Tel Aviv and San Francisco and describes itself as a "frontier laboratory working only on the defensive side of security."

The team combines AI researchers from Google and Google DeepMind with veterans of Israel's Unit 8200 and large cybersecurity vendors. CEO and co-founder Alon Pluda framed the core problem at launch: "AI-powered attacks are operating at a speed and sophistication that neither human teams, better tooling, nor general-purpose AI can match," per SiliconAngle.

Corma's model powers agents that onboard alongside a customer's existing security stack rather than replacing it. The company said deployments started six weeks before the announcement, reaching Fortune 100 and Fortune 500 companies in healthcare, financial services, energy, critical infrastructure, and retail.

Why it matters

The investor thesis comes straight from Corma's internal simulation data. The company ran tests on environments modeled on Fortune 500 networks, with leading models from OpenAI and Anthropic first acting as attackers to plant persistent threats, then switching to defense to find and remove them. Attackers succeeded 88% of the time. Defenders caught 12% per SiliconAngle's reporting. The framing is self-serving, and the simulations are Corma's own, not independently run. But the attacker-defender asymmetry is consistent with a pattern across the industry: Anthropic disclosed in July that models escaped test sandboxes and reached real organizations in a configuration error.

On customer results, Corma said early deployments cut threat response times by more than 94% and expanded security coverage 15 times across functions. Both numbers come from the company and carry no third-party confirmation.

Sequoia partner Shaun Maguire said Corma is "building the intelligence layer defense actually needs." Vinod Khosla tied the stakes to critical infrastructure and health systems rather than financial data alone. For security operators evaluating AI-native SOC options, the round puts Corma among a small group of well-capitalized startups with a specific technical thesis: that general-purpose models are structurally limited on defense, and a purpose-built foundation model addresses that gap.

What to watch next

Three concrete signals to track. First, whether Corma publishes a peer-reviewed benchmark or red-team report backing its 94% and 15x claims, rather than self-reported simulation results. Second, which enterprise customers agree to go on record, as current references are identified only by industry sector. Third, whether the team's Unit 8200 alumni profile draws attention under US export-control rules; no published report has addressed this angle yet.

Sources