Source: SecurityWeek — 2026-08-04
Summary
Obsidian Security raised an $85 million Series D led by Crescent Cove Advisors, with Greylock and Menlo Ventures participating, pushing its valuation past $1.1 billion and total funding to $200 million. The platform governs AI agents as they operate across third-party enterprise SaaS applications — treating agents as identities to be monitored and controlled, not human employees. The company reports over 100 customers spending $100K+ a year, more than 14 spending over $1 million a year, and 60 Fortune 500 customers across financial services, social media, and telecom.
Key Takeaways
- $85 million Series D led by Crescent Cove Advisors, with Greylock and Menlo Ventures also participating, crosses Obsidian past a $1.1 billion valuation.
- Total capital raised to date reaches $200 million across all rounds.
- The product is an identity/access-governance layer purpose-built for non-human AI agent identities connecting into enterprise SaaS apps, distinct from traditional human-focused identity governance.
- Customer traction: 100+ accounts spending $100K+/year, 14+ spending over $1M/year, and 60 Fortune 500 companies as customers.
- Customer base spans financial services, social media, and telecom — sectors with heavy regulatory exposure and high-value data, where uncontrolled agent access carries outsized risk.
- The round signals investor conviction that agent governance/security is becoming its own durable category as enterprises plug AI agents into existing SaaS stacks at scale.
Reel Script
Hook: Your company just gave an AI agent a login to Salesforce, Slack, and your HR system — with almost no one watching what it actually does with that access. That gap just became a $1.1 billion company.
Core Concept: For years, enterprise security has been built around governing human identities: who can log into what, what permissions an employee has, when to revoke access when someone leaves. AI agents break that model because they're not humans — they're software identities that enterprises are now connecting into every SaaS app they own, often with broad, standing permissions and no employee-style offboarding process. Obsidian Security's bet is that agents need the same kind of governance humans get, but built for how agents actually behave: constant, high-volume API calls instead of occasional logins, and access patterns that don't map to a 9-to-5 human workflow. Think of it as the difference between a security guard who checks badges at the door once a day, versus a supervisor who watches a robot on the factory floor continuously, because the robot can touch far more machines, far faster, than any single person could.
Hands-On: Picture the flow: an enterprise app — say, a CRM or a financial system — has an AI agent plugged into it, granted API access to read and write data. Obsidian's governance layer sits between that agent and the app, monitoring exactly what the agent touches and enforcing limits on what it's allowed to do, the same conceptual slot that identity governance tools occupy for human employees, just re-architected for non-human, machine-speed identities. The traction numbers back up that this isn't hypothetical: more than 100 customers already pay Obsidian over $100,000 a year, and more than 14 of those pay over $1 million a year — the kind of spend tier that signals mission-critical infrastructure, not a pilot project. Sixty of those customers are Fortune 500 companies, concentrated in financial services, social media, and telecom — three industries where a rogue or over-permissioned AI agent touching customer data is a regulatory and reputational nightmare. The $85 million Series D, led by Crescent Cove Advisors with Greylock and Menlo Ventures also in, brings total funding to $200 million and values the company above $1.1 billion.
Takeaway: As companies rush to connect AI agents to real business systems, the identity-governance layer for those agents is turning into one of the most concrete, fundable problems in enterprise AI — this isn't speculative infrastructure, it's already at eight-figure customer spend. If your company is granting AI agents access to production systems, ask who is watching what those agents actually do with it.