Oasis Security raises $120 million to manage access for AI agents and service accounts

Oasis Security has raised $120 million from investors including Craft, Sequoia Capital and Accel. The startup builds tools to manage how non-human accounts, from service accounts to AI agents, access corporate systems, and this area is quickly becoming a new front in cybersecurity.

What Oasis does

Oasis Security's product tackles a problem that until recently seemed niche: who gets into a system, and with what permissions, when the entity is a program rather than an employee. Companies have long run service accounts, integrations, bots, automations and internal scripts. Now AI agents are joining them, with access to documents, corporate knowledge bases, email, CRM and other work systems. The more such entities there are, the harder it is to know what permissions they hold and where those permissions become excessive.

The problem is that non-human accounts often fall outside classic security processes. An employee has a manager, a role, a termination date and a clear point at which access can be revoked. For an AI agent or a service account, it is different: access can be granted quickly, used across several systems at once and kept longer than necessary. As a result, companies end up with a growing layer of privileges that nobody is held accountable for as strictly as for ordinary users' access.

Why demand is growing

The market itself is pushing in this direction. Businesses deploy AI tools not only for chat and search but for real actions: reading internal data, launching processes, working with tickets, analytics and customer systems. Each such scenario requires credentials, tokens, roles and access rules. The more companies move from experiments to automation, the more pressing the question becomes: how to control not people but digital workers acting on the company's behalf.

In practice, the risks usually look like this:

• AI agents get access to several internal services and SaaS products at once.
• Service accounts live for years and rarely get the same review as employee access.
• Permissions accumulate as new integrations launch and almost never get rolled back.
• Compromising a single technical account can open a path to a chain of other systems.

That is why non-human identity management is starting to stand out as a separate layer of the security stack. To investors, it looks less like a narrow feature and more like an infrastructure problem that will only grow as autonomous AI scenarios spread. Companies could once tolerate chaos around machine accounts, but with mass adoption of agents, that approach becomes too expensive and risky.

A signal to the market

The size of the round says a lot. $120 million is not a bet on a trendy term but a large investment in a specific pain point of the enterprise segment. The participation of Craft, Sequoia Capital and Accel suggests that big funds see a long growth cycle in this area: companies will not simply buy one more security tool but will rethink the access model itself for programs, bots and AI agents. For Oasis Security, this provides the resources to scale the product quickly and establish itself in a category that is only now taking shape.

The deal also shows a shift within the cybersecurity market itself. A few years ago, the main conversation centered on protecting devices, networks and employees. Now the focus is moving to digital identity without a human: who launches an action, on the basis of which permissions, and how to verify it at any moment. As AI moves from demos into working processes, such platforms may become a mandatory element of corporate architecture.

What this means

The Oasis Security story confirms a simple conclusion: the more AI agents and automations a business has, the more important it is to manage their access as strictly as employee access. The security market is already starting to reorganize around this reality, and large investments show that demand for such solutions will grow along with corporate AI.