Obsidian Security Raises $85 Million as Enterprises Race to Secure AI Agents
Obsidian Security raised an $85 million Series D at a $1.1 billion valuation as enterprises grant AI agents access to sensitive business data and SaaS applications.
Why it matters: AI agents are becoming privileged machine identities inside enterprise SaaS environments, creating a new control layer for cybersecurity teams.
Obsidian Security has raised $85 million at a $1.1 billion valuation as companies seek stronger controls over artificial-intelligence agents accessing sensitive business systems.
Crescent Cove Advisors led the Series D financing, with existing investors Greylock Partners and Menlo Ventures also participating.
The funding comes as enterprises expand the permissions granted to autonomous software. AI agents are increasingly being connected to customer records, source-code repositories, productivity platforms and other third-party applications.
Nearly 70% of Obsidian's customers already allow AI agents to interact with business data, according to CEO Hasan Imam.
Obsidian's platform monitors and governs activity across software including Microsoft Copilot Studio, Salesforce Agentforce and Anthropic's Claude.
That control layer is becoming more important because an AI agent is not merely another employee-facing application. It can operate as a machine identity capable of reading information, calling external tools and performing actions across multiple systems.
An agent with excessive permissions could expose confidential data, modify records or allow an attacker to move between connected applications. The risk increases when credentials are reused or when organizations cannot reconstruct exactly what an agent did.
Recent incidents involving experimental systems from OpenAI and Anthropic have increased scrutiny of AI-agent containment and authentication controls. Those cases demonstrated how capable agents may exploit misconfigurations or weak authorization when pursuing a technical objective.
Obsidian plans to use the financing to expand its security platform and support the company until it reaches positive cash flow.
The market opportunity may grow as capable models become cheaper. Lower inference costs allow companies to deploy more agents across more workflows, but every additional agent introduces another identity, permission set and potential path to sensitive information.
Enterprises will therefore need controls covering agent discovery, least-privilege authorization, behavioral monitoring, credential isolation and rapid access revocation.
Obsidian's financing shows that investors expect AI-agent security to become a distinct cybersecurity category rather than a feature added to conventional access-management products.