Shai Morag has raised $60 million for Oak, an identity security startup built around a simple problem enterprises are only starting to face: AI agents now need access, and most companies don't know who is watching it.
Shai Morag has already sold three cybersecurity companies for a combined sum that tops half a billion dollars. That's not a rookie founder. On July 15 he came out of stealth with Oak, and the timing tells you where enterprise security is moving.
Oak is the identity security startup Morag built with co-founder Tal Marom. It's headquartered in Tel Aviv and San Francisco, with a $60 million seed round behind it: co-led by Accel, Greylock Partners and CRV, with Hetz Ventures, AlphaDrive Ventures and strategic angels also joining. The product isn't a prototype. It's already generally available. Enterprise customers are running it live, though Oak hasn't named them.
That silence is worth keeping in the story. A young security company can say large enterprises are using its platform. Until it names them, you should treat the customer claim as real but incomplete.
Morag knows this market cold. This is his fourth cybersecurity company. He sold Integrity-Project to Mellanox in 2014 and Secdo to Palo Alto Networks in 2018. Then he built Ermetic, a cloud identity security company, and sold it to Tenable in 2023 for $265 million, according to Calcalist. He stayed on as a senior product executive at Tenable before leaving to start Oak with Marom, who had led product teams at Tenable and Salesforce.
By the fourth company, the bet is no longer theory. Morag's bet is that identity, not malware or perimeter defense, is where enterprises are actually exposed, and that the tools built to manage it were never designed to talk to each other. "The tools enterprises rely on today were never designed to work together," he told Calcalist.
Oak calls its product an AI native Identity Operating System. The phrase is doing some marketing work. The mechanics underneath are more useful: a live identity graph built from raw evidence pulled out of on-prem systems, cloud infrastructure, SaaS tools and homegrown applications. Oak says that graph tracks every identity in an organization, whether it belongs to a person, a machine or an AI agent, under one control plane.
The AI agent no one offboards
That last category is the reason this round closed fast. Plug an AI agent into your internal systems and it gets credentials, API keys and permissions just like a new hire would. Nobody holds an onboarding meeting for it. Nobody runs an exit interview when its task is done.
According to TechCrunch, Oak's team spent months talking to more than 100 CISOs and identity and access management leaders before writing a line of the product. That is the right order. If you're building for security buyers, you don't begin with a grand theory about agents. You begin with the broken workflow they already hate.
Morag described that workflow plainly to TechCrunch: "Right now, the whole process is too manual, and it's operations-based, not risk-based, for instance, there's no trigger when an employee logs in from an unusual location."
That gap isn't hypothetical. It's the daily condition of identity governance in many large companies, where quarterly certifications, IT tickets and stale permissions have become a ritual. Access gets approved. Access stays. Months later, nobody is entirely sure whether it still maps to real work.
Oak wants to replace that cycle with continuous remediation. Its system watches how access is actually used and removes permissions when they stop matching real activity. That sounds obvious only if you've never watched a large company try to unwind years of accumulated identity sprawl.
The money is betting on Morag's read
Accel, Greylock and CRV don't co-lead a $60 million seed round on language alone. Three firms of that size signing onto one round, with Hetz Ventures and AlphaDrive Ventures beside them, is a bet on Morag's read of the market as much as on the product Oak has shipped.
Calcalist reported that Oak was founded in December 2025 and built a 50-person team while still in stealth, with most of that headcount expected to shift toward the U.S. going forward. Fifty people, built entirely in stealth, is a serious operation for a company that hadn't even launched yet. It suggests customers, product proof and investor conviction existed before this week's announcement.
The harder question is what happens when AI agents move from experiment to infrastructure. One agent with a handful of permissions is manageable. A thousand agents touching finance systems, customer databases, code repositories and support tools is something else entirely.
What Oak hasn't disclosed is which enterprises are live on the platform, or how its identity graph performs when an organization's AI agent count climbs into the thousands. That's the number worth watching. Agentic AI isn't adding a few new accounts to an already messy identity stack. It's multiplying the stack before most companies have finished cleaning up the last one.
Also read: Sequoia Capital Pours $45 Million Into an AI Startup That Calls Its Product an Employee, A Former Red Bull F1 Engineer Just Raised Germany's Biggest Seed Round Ever, Scribble Network: The AI Visibility Platform for Brands
