Oak Raises $60 Million to Solve the Identity Crisis That AI Agents Are Making Worse

What happens when a company has 5,000 human employees and 10,000 AI agents, all needing access to different systems, databases, and APIs? That question kept Shai Morag up at night. The serial entrepreneur who sold Ermetic to Tenable for $265 million in 2023 just emerged from stealth with his new company Oak, armed with a $60 million seed round from Accel, CRV, and Greylock, to build the identity infrastructure that the age of AI agents demands. The round is one of the largest seed financings ever raised by an Israeli startup, and it signals something bigger: identity management for non-human actors is becoming critical infrastructure, and fast.

The problem Oak is tackling is straightforward in concept but terrifying in complexity. Legacy identity and access management systems were built for a world where only humans needed permissions. Today, AI agents operate autonomously across cloud environments, accessing databases, running code, and making API calls. Those agents need granular, auditable, and revocable access, but most organizations manage them with the same blunt tools they used for human employees. The result is a gaping security hole that grows wider every time a company deploys another agent.

The $60 Million Bet on AI-Native Identity

Oak's $60 million seed round is eye-catching not just for its size, but for who is writing the checks. Accel, CRV, and Greylock co-led the round, three venture firms with deep cybersecurity portfolios. Accel partner Andrei Brasoveanu, who led Accel's investment in Ermetic when it was still pre-revenue, said Morag had an informal standing offer from Accel to back whatever he built next. That kind of follow-on conviction from a top-tier VC is rare, and it speaks to Morag's track record of building exactly what the market needs next.

Morag spent two decades in cybersecurity, including a stint as an army major in Israel, and had three exits before founding Oak. After Ermetic's acquisition, he stayed at Tenable as CPO, but when Tenable CEO Amit Yoran became ill, Morag decided the moment was right to build something new rather than step into a leadership role at the public company. He co-founded Oak with Tal Marom, a product leader he met at Tenable who previously held senior product roles at Salesforce and in the Israeli military's technology units.

Before building anything, Oak's founding team spent months talking to 100 CISOs and IAM leaders to understand the exact pain points. What they found was a market crying out for consolidation: most enterprises manage human identity, machine identity, and AI agent identity with three separate toolsets that do not talk to each other. Oak's solution is a unified control plane that governs identity across every type of actor in an organization.

What Oak Actually Builds

Oak describes itself as an AI-native identity management platform. That label carries weight. The startup built an AI connector framework that maps every permission to actual application usage in real time, removing credentials that are no longer needed as soon as they become stale. This is a significant upgrade from the legacy approach of periodic access reviews, which most security teams run quarterly or annually, leaving months of exposure windows open.

The platform also detects anomalous access patterns automatically. Morag described the current state of identity management as too manual and operations-based rather than risk-based. In a typical enterprise today, there is no automated trigger when an employee or AI agent logs in from an unusual location or requests access to a system they have never touched before. Oak aims to close that gap by treating identity as a continuous, risk-scored signal rather than a static permission set that gets reviewed once a quarter.

For founders building AI agents, this has immediate implications. Every agent you deploy needs an identity with scoped permissions, and current tools make it far too easy to accidentally give an agent more access than it needs. The principle of least privilege, a cornerstone of security engineering, becomes exponentially harder to enforce when the number of entities in your system can double or triple overnight as agents scale.

Why This Matters for AI-First Startups

Oak's emergence is another data point in a pattern that is becoming increasingly clear: the AI agent infrastructure stack is forming right now, and it will look very different from the SaaS stack of the last decade. Identity management is the second layer of that stack, sitting just above the compute and model layers. Every startup building AI agents today should be asking whether their identity architecture will scale to a world where non-human actors outnumber human ones ten to one.

The security implications go beyond convenience. Outdated credentials and poor IAM are already among the most common attack vectors in enterprise security breaches. AI agents multiply that attack surface dramatically because each agent represents a new potential entry point, often with automated access to sensitive systems. If an attacker compromises an agent's credentials, they inherit every permission that agent has, which could include database access, API keys, and deployment pipelines.

Oak's $60 million seed validates a market thesis that multiple VCs are now racing to fund: the AI agent infrastructure layer needs entirely new tooling, and the companies that build it will be as important as the companies building the agents themselves. Morag told TechCrunch that VC interest was strong from the outset, and with good reason. The identity management market is a multi-billion dollar category that is about to be disrupted by the very technology that is straining it.

The Competitive Landscape and What Comes Next

Oak will not have the market to itself. The traditional IAM vendors, including Okta, Microsoft, and Ping Identity, are all adding AI agent capabilities to their existing platforms. But Oak is betting that incumbents will struggle to adapt legacy architectures to a paradigm where non-human actors are the primary users. Building AI-native from scratch, rather than bolting agent support onto a human-first system, could be a decisive advantage.

Morag said Oak already has its product generally available and deployed by enterprise clients, though the company has not disclosed which ones. The team of 50 people is actively hiring, particularly in the US, where a majority of staff will soon be based. Oak is investing heavily in R&D and growth, with Morag telling his wife this will be his last company and that he will go big or go home.

For founders, the takeaway is clear. Identity management for AI agents is not a future problem. Enterprises are buying solutions for it today, and if Oak executes on its vision, it could become the standard for how organizations govern access in an agent-filled world. The question every AI founder should be asking is not whether their agents need identity management, but which provider will manage it when they have 100 agents, then 1,000, then 10,000.