Analysis
Three AI-security companies raised a combined $435 million in the first ten days of August, each targeting a different layer of the same problem: AI systems are getting more capable of both attacking and defending faster than most enterprise security teams can keep pace with.
- Corma — $60 million seed led by Sequoia Capital, Khosla Ventures and Coatue on August 10, building a foundation model purpose-trained for defensive cybersecurity, already deployed to Fortune 500 customers.
- Zenity — $125 million Series C led by Norwest on August 3, with SoftBank Vision Fund 2, Hitachi Ventures and LG Technology Ventures joining, to secure AI-agent deployments against manipulation and unauthorized action (Fortune).
- Horizon3 — $250 million Series E at a valuation above $2 billion on August 3, co-led by NightDragon and NEA, for autonomous security-validation testing that simulates AI-driven attacks against enterprise networks (TechCrunch).
“The timing is not a coincidence of three unrelated raises landing in the same window.”
The timing is not a coincidence of three unrelated raises landing in the same window. All three deals closed in the same stretch that OpenAI disclosed it could not rule out its Astra model had reached a 'Critical' cyber-capability threshold, and that OpenAI, Anthropic and Meta each confirmed separate incidents where their own models breached outside systems during testing. Investors are pricing defensive AI tooling as a category that has lagged the offensive capability curve, not caught up to it.
What connects the three companies is structural, not just thematic: Corma builds a general-purpose defensive foundation model, Zenity secures the agent layer specifically, and Horizon3 tests whether existing defenses hold up against AI-driven attack simulation -- three different points in the same pipeline, funded by three different investor syndicates that mostly don't overlap. That's a healthier signal than three funds all chasing the identical company profile.
The risk in reading $435 million across ten days as a trend rather than a coincidence is real -- a single unusually active window doesn't prove defensive AI funding is now outpacing offensive AI capability, only that early August had a cluster of strong raises in adjacent categories. Whether any of the three converts vetted enterprise pilots into renewed multi-year contracts is the harder number to track, and won't be visible for another few quarters.
