Hush Security Raises $30 Million to Put AI Agents Under Proper Control

Key points

• Hush Security closed a $30 million Series A round on 9 July 2025, bringing its total funding to $41 million.
• Akamai Technologies led the round alongside existing backers Battery Ventures and YL Ventures.
• Founded in 2024, the company came out of stealth in September 2025 with a platform designed to track and control AI agents inside corporate networks.
• The platform issues short-lived, task-specific permissions to AI agents rather than permanent passwords, and logs every action they take.
• Hush plans to use the money to hire engineers and salespeople and deepen support for identity and access management ecosystems.

Companies today run AI agents, programs that can book meetings, query databases, write code, or move files without a human clicking anything. Most businesses have no clear picture of what those agents are doing or what sensitive systems they can reach. That gap is exactly what Hush Security is selling against.

Hush Security, a startup founded in 2024 and based in Tel Aviv, announced a $30 million Series A funding round backed by internet infrastructure company Akamai Technologies and returning investors Battery Ventures and YL Ventures. Total funding now stands at $41 million.

What does Hush Security's platform actually do?

In plain terms: it gives every AI agent an identity card, watches what the agent does, and can cut its access instantly if something goes wrong.

Instead of handing an AI agent a permanent password, or API key, that lets it roam freely, Hush issues what it calls just-in-time permissions. These are single-use visitor passes: the agent gets access to exactly one resource, for exactly as long as the task takes, and nothing more. When the task's done, the pass expires.

Every action the agent takes is written to an audit trail, a running log that compliance teams or investigators can review later. A central kill switch lets administrators shut down any agent across the whole organisation at once.

The company also maps which tools and external services each agent can reach, including MCPs, which stands for Model Context Protocol connectors, the plug-in-style links that let AI agents talk to outside software and data sources.

Why does this matter to ordinary employees?

If your company uses AI tools that can read emails or approve transactions, those tools carry real risk when they're not tracked. A misbehaving or hijacked agent could quietly leak sensitive data or make changes nobody authorised.

Hush argues that businesses know how to manage access for human staff but haven't applied the same discipline to software that acts on its own. CEO Micha Rave said: "AI agents need strict identity, not just API keys."

The funding will go toward hiring, building out integrations with identity and access management (IAM) systems, the software platforms companies use to control who, or what, can log in to which systems, and expanding partnerships. Our piece on Oak's $60 million raise on 16 July shows how much capital is now chasing this same non-human identity problem. The vendor field is filling fast, and buyers will soon have to pick.

Common questions

Does this affect regular employees or just IT teams?

Mostly IT and security teams will configure the platform, but employees who rely on AI tools at work benefit indirectly because properly governed agents are less likely to expose the personal or business data those tools handle.

Is AI agent security a new regulatory concern?

Regulators haven't issued final rules specific to AI agents, though guidance from bodies such as the U.S. Securities and Exchange Commission on material cyber risk disclosure, under its 2023 final rule, already covers incidents involving automated systems that cause significant harm.