Tel Aviv and San Francisco startup Oak came out of stealth this week with $60 million in seed funding to build an AI-native identity operating system, a bet that identity security must be rebuilt for a world run by autonomous agents. Accel, Greylock and CRV co-led the round, and Hetz Ventures and AlphaDrive Ventures joined them.
The increase is important because it points to a problem that founders already feel. Each new agent, script, and AI service account is another identity that requires permissions, and older tools were designed for a slower, human-only enterprise. So as teams add automation, the security surface area increases faster than most people can keep up.
What the oak builds
Oak describes its product as a single control plane that governs people, machines and AI agents together. Instead of consolidating existing systems, the company wants a single place to manage identities, access and permissions. It’s an accurate reading of where enterprise security is heading.
The timing details are revealing. According to the company, Oak closed the cycle late last year, then waited to go public until the product was generally available and already rolling out to paying enterprise customers. Founders can take this move, because arriving with proof is better arriving with a pitch.
The founding team also brings in previous exits, which is why a seed round grew to this size. Investors don’t fund an idea here. They finance the execution against a problem with the actual budgets attached.
Why Agent Sprawl breaks old tools
For years, security spending has focused on the perimeter. This model comes under severe strain when the number of players within a company multiplies. As trade agent and automated workflows spread, machines demand access on a scale that no human repository was designed to handle.
The risk is discreet but costly. An agent with excessive permissions is a violation waiting to happen, and most early teams cannot see all the credentials they have issued. In other words, permissions, not passwords, become the first line.
Oak argues that identity should be treated as infrastructure, not paperwork. When access control is a system rather than a spreadsheet, it grows with the business instead of breaking under pressure.
The scope is less relevant now. The real question is who and what is allowed to act inside your systems.
What Founders Should Do This Quarter
You don’t need a $60 million platform to take identity seriously. Start by listing all non-human identities used by your product, including API keys, service accounts, and any AI agents that have write access. Most teams are surprised by the count.
Then apply least privilege. Grant each agent the narrowest permissions that allow it to function and rotate credentials on a set schedule. These habits cost little, but they prevent the kind of incident that can damage the reputation of a young company.
Finally, consider access control as part of your AI infrastructure from day one. It’s much harder to modernize identity after you scale than to build it from the start, so the cheapest time to act is now.
The signal behind the trick
Oak’s funding is another sign that security dollars are moving toward orchestration and control of policies in agent-heavy environments. Investors want the tools that decide access, just like they argued compliance software which integrates rules into automated work.
You can see the same thesis at Oak identity operating system positioning, which defines identity as a basic infrastructure. For the founders, the conclusion is straightforward, as AI adoption and identity risk grow together.
The oak seed round at a glance
| Detail | Figure |
|---|---|
| Amount raised | 60 million dollars |
| Financing stage | Seed |
| Co-responsible | Acceleration, Greylock, CRV |
| Headquarters | Tel Aviv and San Francisco |
Identity Security Questions Asked by Founders
What is an AI-native identity operating system? It is a single system that manages access for people, machines and AI agents together, instead of separate existing tools linked together.
Why is AI making identity more difficult? Each agent is a new identity with its own permissions, so the number of actors needing access grows quickly and becomes difficult to monitor.
