Oak raises a massive $60M seed to police AI agents

Signals Inbox·July 15, 2026·Cybersecurity

Oak has emerged from stealth with a $60 million seed to rebuild enterprise identity around AI agents. The round is unusually large, but the sharper signal is timing: companies are giving autonomous software access to sensitive tools faster than their old identity systems can track or restrict it.

Q1What actually happened?

According to Oak’s official announcement, the company emerged from stealth with a $60 million seed led by Accel, Greylock, and CRV. Its product is already generally available and being used by enterprise customers.

Q2Why is a $60 million seed so unusual?

Because Oak was founded only in December 2025, and $60 million is one of the largest seed rounds ever raised by an Israeli cybersecurity startup. Oak already built a team of around 50 people during stealth, so this looks less like a normal seed experiment and more like a company trying to start at scale.

Q3What identity problem are AI agents creating?

Normal identity tools were built mainly for employees logging into apps. AI agents can use several systems, move data, call other agents, and take actions without a person clicking each button. Many still inherit broad human permissions, making it difficult to know what they can access, what they actually use, or when that access should disappear.

Q4What does Oak do differently?

Oak connects to cloud, on-premise, SaaS, and custom applications, then builds a live map of every human, machine, and AI identity. It compares the permissions each identity holds with what it really uses, flags unnecessary access, and can remove that access continuously instead of waiting for a quarterly review.

Q5Is Oak alone in chasing this market?

Not at all. Arcade.dev raised $60 million in June 2026 to control which actions AI agents can take. Saviynt raised $700 million at a roughly $3 billion valuation to manage human and non-human identities. Large security companies are also expanding into identity. Oak is entering a crowded race, but the repeated mega-rounds suggest the problem is becoming real very quickly.

Q6What is the real point of tension?

Companies want agents that can open files, update databases, approve requests, and run workflows. But every useful action requires access, and every extra permission creates another way to leak data or cause damage. The agent race could therefore be slowed not by model intelligence, but by whether security teams can safely give software the keys.

Q7So should I care?

Yes, because identity may become the control layer that decides which AI agents enterprises can actually deploy. Oak still needs to prove that one platform can replace deeply embedded tools and win large customers. But a $60 million seed, an experienced founding team, and an already live product show investors expect agent identity to become a major security category, not a small feature.